{ "generated": "2026-10-12T09:47:32Z", "schema_version": "2.0", "sources": { "openrouter": "https://openrouter.ai/api/v1/models", "artificial_analysis": "https://artificialanalysis.ai/api/v2/data/llms/models", "pinchbench": "https://pinchbench.com", "arena": "https://arena.ai/leaderboard/text", "eq_bench": "https://eqbench.com" }, "score_methodology": { "reasoning": "Weighted average of: AA Intelligence Index (3x), GPQA (2.5x), MMLU-Pro (2x), HLE (1.5x), AIME 2025 (1x). Scale 0-100.", "coding": "Weighted average of: AA Coding Index (3x), LiveCodeBench (2x), TerminalBench Hard (2x), SciCode (1x). Scale 0-100.", "agentic": "Weighted average of: PinchBench Best (4x), IFBench (3x), PinchBench Avg (2x). Scale 0-100.", "blended_cost": "3:1 input-to-output ratio: (3 * input_per_m + output_per_m) / 4" }, "models": [ { "id": "anthropic/claude-fable-5", "canonical_slug": "anthropic/claude-5-fable-20260609", "name": "Claude Fable 5", "provider": "Anthropic", "description": "Claude Fable 5 is a Mythos-class model from Anthropic, built for autonomous knowledge work and coding. It supports text, image, and file inputs with text output, with reasoning support and...", "context_window": 1000000, "max_output": 128000, "modalities": { "input": [ "text", "image", "file" ], "output": [ "text" ] }, "tokenizer": "Claude", "pricing": { "input_per_m": 10.0, "output_per_m": 50.0, "cache_read_per_m": 1.0, "cache_write_per_m": 12.5, "web_search_per_req": 0.01, "blended_per_m": 20.0 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": true, "structured_output": true }, "endpoint_stats": { "providers": [ "Amazon Bedrock", "Anthropic", "Azure", "Claude Platform on AWS", "Google" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "claude-fable-5", "public_rubric_0_100": 82.6, "public_elo_norm": 2049.7, "public_traits_17": { "demonstrated_empathy": 16.12, "pragmatic_ei": 16.0, "depth_of_insight": 18.0, "social_dexterity": 15.15, "emotional_reasoning": 17.16, "message_tailoring": 14.69, "boundary_setting": 14.54, "safety_conscious": 13.62, "moralising": 7.62, "compliant": 9.73, "challenging": 13.35, "warmth": 14.04, "validating": 13.35, "analytical": 18.04, "reactive": 9.0, "conversational": 11.23, "humanlike": 15.27 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1781007515, "last_updated": "2026-10-12T09:47:31Z" }, { "id": "anthropic/claude-haiku-4.5", "canonical_slug": "anthropic/claude-4.5-haiku-20251001", "name": "Claude Haiku 4.5", "provider": "Anthropic", "description": "Claude Haiku 4.5 is Anthropic\u2019s fastest and most efficient model, delivering near-frontier intelligence at a fraction of the cost and latency of larger Claude models. Matching Claude Sonnet 4\u2019s performance...", "context_window": 200000, "max_output": 64000, "modalities": { "input": [ "text", "image", "file" ], "output": [ "text" ] }, "tokenizer": "Claude", "pricing": { "input_per_m": 1.0, "output_per_m": 5.0, "cache_read_per_m": 0.1, "cache_write_per_m": 1.25, "web_search_per_req": 0.01, "blended_per_m": 2.0 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": true, "structured_output": true }, "endpoint_stats": { "providers": [ "Amazon Bedrock", "Anthropic", "Azure", "Google" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": { "reasoning": 31, "coding": 35.0, "agentic": 85.7 }, "benchmarks": { "artificial_analysis": { "intelligence_index": 31.1, "coding_index": 29.6, "gpqa": 0.646, "hle": 0.043, "scicode": 0.344, "ifbench": 0.42, "terminalbench_hard": 0.273, "livecodebench": 0.511, "aime_25": 0.39, "mmlu_pro": 0.8 }, "eq_bench": { "v3_score": 68.65, "v3_traits": { "analytical": 18.0, "boundary_setting": 13.2, "challenging": 13.3, "compliant": 8.0, "conversational": 11.4, "correctness": 13.6, "demonstrated_empathy": 14.6, "depth_of_insight": 15.6, "emotional_reasoning": 14.9, "humanlike": 14.3, "intellectual_grounding": 12.0, "message_tailoring": 11.5, "moralising": 6.3, "pragmatic_ei": 12.1, "reactive": 10.2, "safety_conscious": 13.2, "social_dexterity": 11.0, "subtext_identification": 15.6, "sycophantic": 2.9, "theory_of_mind": 14.5, "validating": 12.0, "warmth": 11.2 } }, "pinchbench": { "best_score": 89.5, "avg_score": 78.1 }, "arena": { "elo": 1407 } }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false }, "created": 1760547638, "last_updated": "2026-10-12T09:47:31Z", "speed": { "output_tokens_per_sec": 93, "ttft_seconds": 0.4, "source": "artificial_analysis" } }, { "id": "anthropic/claude-opus-4.5", "canonical_slug": "anthropic/claude-4.5-opus-20251124", "name": "Claude Opus 4.5", "provider": "Anthropic", "description": "Claude Opus 4.5 is Anthropic\u2019s frontier reasoning model optimized for complex software engineering, agentic workflows, and long-horizon computer use. It offers strong multimodal capabilities, competitive performance across real-world coding and...", "context_window": 200000, "max_output": 64000, "modalities": { "input": [ "file", "image", "text" ], "output": [ "text" ] }, "tokenizer": "Claude", "pricing": { "input_per_m": 5.0, "output_per_m": 25.0, "cache_read_per_m": 0.5, "cache_write_per_m": 6.25, "web_search_per_req": 0.01, "blended_per_m": 10.0 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": true, "structured_output": true }, "endpoint_stats": { "providers": [ "Amazon Bedrock", "Anthropic", "Azure", "Claude Platform on AWS", "Google" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "claude-opus-4-5-20251101", "public_rubric_0_100": 76.8, "public_elo_norm": 1545.7, "public_traits_17": { "demonstrated_empathy": 15.35, "pragmatic_ei": 15.04, "depth_of_insight": 16.76, "social_dexterity": 14.27, "emotional_reasoning": 16.09, "message_tailoring": 13.62, "boundary_setting": 13.15, "safety_conscious": 12.88, "moralising": 7.77, "compliant": 10.38, "challenging": 12.54, "warmth": 13.62, "validating": 13.12, "analytical": 16.73, "reactive": 9.42, "conversational": 11.85, "humanlike": 14.65 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1764010580, "last_updated": "2026-10-12T09:47:31Z" }, { "id": "anthropic/claude-opus-4.6", "canonical_slug": "anthropic/claude-4.6-opus-20260205", "name": "Claude Opus 4.6", "provider": "Anthropic", "description": "Opus 4.6 is Anthropic\u2019s strongest model for coding and long-running professional tasks. It is built for agents that operate across entire workflows rather than single prompts, making it especially effective...", "context_window": 1000000, "max_output": 128000, "modalities": { "input": [ "text", "image", "file" ], "output": [ "text" ] }, "tokenizer": "Claude", "pricing": { "input_per_m": 5.0, "output_per_m": 25.0, "cache_read_per_m": 0.5, "cache_write_per_m": 6.25, "web_search_per_req": 0.01, "blended_per_m": 10.0 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": true, "structured_output": true }, "endpoint_stats": { "providers": [ "Amazon Bedrock", "Anthropic", "Azure", "Claude Platform on AWS", "Google" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": { "reasoning": 53.9, "coding": 47.6, "agentic": 89.9 }, "benchmarks": { "artificial_analysis": { "intelligence_index": 46.5, "coding_index": 47.6, "gpqa": 0.84, "hle": 0.186, "scicode": 0.457, "ifbench": 0.446, "terminalbench_hard": 0.485 }, "eq_bench": { "elo": 1857.8, "humanlike": 6.9, "safety": 7.0, "assertive": 7.3, "social_iq": 7.5, "warm": 7.1, "analytic": 8.8, "insight": 8.9, "empathy": 8.0, "compliant": 4.8, "moralising": 4.7, "pragmatic": 8.0, "v3_score": 71.85, "v3_traits": { "analytical": 16.7, "boundary_setting": 11.7, "challenging": 12.4, "compliant": 8.2, "conversational": 12.0, "correctness": 14.3, "demonstrated_empathy": 14.9, "depth_of_insight": 15.6, "emotional_reasoning": 14.6, "humanlike": 13.8, "intellectual_grounding": 13.0, "message_tailoring": 13.7, "moralising": 9.4, "pragmatic_ei": 13.9, "reactive": 8.6, "safety_conscious": 12.1, "social_dexterity": 13.5, "subtext_identification": 15.3, "sycophantic": 4.0, "theory_of_mind": 13.5, "validating": 13.2, "warmth": 13.6 }, "public_source_key": "claude-opus-4-6", "public_rubric_0_100": 79.2, "public_elo_norm": 1717.4, "public_traits_17": { "demonstrated_empathy": 15.62, "pragmatic_ei": 15.73, "depth_of_insight": 17.33, "social_dexterity": 14.73, "emotional_reasoning": 16.62, "message_tailoring": 14.23, "boundary_setting": 14.27, "safety_conscious": 13.62, "moralising": 8.81, "compliant": 9.85, "challenging": 13.81, "warmth": 14.04, "validating": 13.38, "analytical": 17.42, "reactive": 9.42, "conversational": 11.58, "humanlike": 14.69 } }, "pinchbench": { "best_score": 93.3, "avg_score": 83.1, "runs": 19 }, "arena": { "elo": 1499, "rank": 2, "category": "text" } }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": true, "eq_bench_public": true }, "created": 1770219050, "last_updated": "2026-10-12T09:47:31Z", "speed": { "output_tokens_per_sec": 46.1, "ttft_seconds": 1.655, "source": "artificial_analysis" } }, { "id": "anthropic/claude-opus-4.7", "canonical_slug": "anthropic/claude-4.7-opus-20260416", "name": "Claude Opus 4.7", "provider": "Anthropic", "description": "Opus 4.7 is the next generation of Anthropic's Opus family, built for long-running, asynchronous agents. Building on the coding and agentic strengths of Opus 4.6, it delivers stronger performance on...", "context_window": 1000000, "max_output": 128000, "modalities": { "input": [ "text", "image", "file" ], "output": [ "text" ] }, "tokenizer": "Claude", "pricing": { "input_per_m": 5.0, "output_per_m": 25.0, "cache_read_per_m": 0.5, "cache_write_per_m": 6.25, "web_search_per_req": 0.01, "blended_per_m": 10.0 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": true, "structured_output": true }, "endpoint_stats": { "providers": [ "Amazon Bedrock", "Anthropic", "Azure", "Claude Platform on AWS", "Google" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "claude-opus-4-7", "public_rubric_0_100": 82.6, "public_elo_norm": 1884.3, "public_traits_17": { "demonstrated_empathy": 16.42, "pragmatic_ei": 16.85, "depth_of_insight": 17.76, "social_dexterity": 15.81, "emotional_reasoning": 17.04, "message_tailoring": 15.35, "boundary_setting": 14.62, "safety_conscious": 13.15, "moralising": 6.88, "compliant": 9.46, "challenging": 13.23, "warmth": 13.46, "validating": 13.46, "analytical": 17.81, "reactive": 8.46, "conversational": 12.77, "humanlike": 15.92 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1776351100, "last_updated": "2026-10-12T09:47:31Z" }, { "id": "anthropic/claude-opus-4.8", "canonical_slug": "anthropic/claude-4.8-opus-20260528", "name": "Claude Opus 4.8", "provider": "Anthropic", "description": "Claude Opus 4.8 is Anthropic's most capable generally available model in the Opus family. It supports text, image, and file inputs with text output, with reasoning support and a 1M-token...", "context_window": 1000000, "max_output": 128000, "modalities": { "input": [ "text", "image", "file" ], "output": [ "text" ] }, "tokenizer": "Claude", "pricing": { "input_per_m": 5.0, "output_per_m": 25.0, "cache_read_per_m": 0.5, "cache_write_per_m": 6.25, "web_search_per_req": 0.01, "blended_per_m": 10.0 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": true, "structured_output": true }, "endpoint_stats": { "providers": [ "Amazon Bedrock", "Anthropic", "Azure", "Claude Platform on AWS", "Google" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "claude-opus-4-8", "public_rubric_0_100": 83.3, "public_elo_norm": 2029.8, "public_traits_17": { "demonstrated_empathy": 16.58, "pragmatic_ei": 16.54, "depth_of_insight": 17.8, "social_dexterity": 15.88, "emotional_reasoning": 17.18, "message_tailoring": 15.42, "boundary_setting": 14.27, "safety_conscious": 13.5, "moralising": 6.88, "compliant": 9.42, "challenging": 13.35, "warmth": 14.38, "validating": 13.58, "analytical": 17.46, "reactive": 8.62, "conversational": 13.15, "humanlike": 15.73 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1779905091, "last_updated": "2026-10-12T09:47:31Z" }, { "id": "anthropic/claude-sonnet-4.5", "canonical_slug": "anthropic/claude-4.5-sonnet-20250929", "name": "Claude Sonnet 4.5", "provider": "Anthropic", "description": "Claude Sonnet 4.5 is Anthropic\u2019s most advanced Sonnet model to date, optimized for real-world agents and coding workflows. It delivers state-of-the-art performance on coding benchmarks such as SWE-bench Verified, with...", "context_window": 1000000, "max_output": 64000, "modalities": { "input": [ "text", "image", "file" ], "output": [ "text" ] }, "tokenizer": "Claude", "pricing": { "input_per_m": 3.0, "output_per_m": 15.0, "cache_read_per_m": 0.3, "cache_write_per_m": 3.75, "web_search_per_req": 0.01, "blended_per_m": 6.0 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": true, "structured_output": true }, "endpoint_stats": { "providers": [ "Amazon Bedrock", "Anthropic", "Azure", "Claude Platform on AWS", "Google" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "claude-sonnet-4.5", "public_rubric_0_100": 72.2, "public_elo_norm": 1511.0, "public_traits_17": { "demonstrated_empathy": 13.92, "pragmatic_ei": 13.31, "depth_of_insight": 16.67, "social_dexterity": 11.92, "emotional_reasoning": 15.42, "message_tailoring": 11.69, "boundary_setting": 14.19, "safety_conscious": 12.85, "moralising": 10.65, "compliant": 8.92, "challenging": 14.15, "warmth": 11.77, "validating": 11.35, "analytical": 17.15, "reactive": 11.08, "conversational": 10.42, "humanlike": 13.73 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1759161676, "last_updated": "2026-10-12T09:47:31Z" }, { "id": "anthropic/claude-sonnet-4.6", "canonical_slug": "anthropic/claude-4.6-sonnet-20260217", "name": "Claude Sonnet 4.6", "provider": "Anthropic", "description": "Sonnet 4.6 is Anthropic's most capable Sonnet-class model yet, with frontier performance across coding, agents, and professional work. It excels at iterative development, complex codebase navigation, end-to-end project management with...", "context_window": 1000000, "max_output": 128000, "modalities": { "input": [ "text", "image", "file" ], "output": [ "text" ] }, "tokenizer": "Claude", "pricing": { "input_per_m": 3.0, "output_per_m": 15.0, "cache_read_per_m": 0.3, "cache_write_per_m": 3.75, "web_search_per_req": 0.01, "blended_per_m": 6.0 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": true, "structured_output": true }, "endpoint_stats": { "providers": [ "Amazon Bedrock", "Anthropic", "Azure", "Claude Platform on AWS", "Google" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": { "reasoning": 50.4, "coding": 46.4, "agentic": 85.7 }, "benchmarks": { "artificial_analysis": { "intelligence_index": 44.4, "coding_index": 46.4, "gpqa": 0.799, "hle": 0.132, "scicode": 0.469, "ifbench": 0.412, "terminalbench_hard": 0.462 }, "eq_bench": { "elo": 1876.8, "humanlike": 7.1, "safety": 6.7, "assertive": 6.6, "social_iq": 7.8, "warm": 6.9, "analytic": 8.8, "insight": 8.9, "empathy": 8.5, "compliant": 4.8, "moralising": 2.9, "pragmatic": 8.3, "v3_score": 71.7, "v3_traits": { "analytical": 16.7, "boundary_setting": 11.6, "challenging": 11.8, "compliant": 7.8, "conversational": 11.8, "correctness": 14.4, "demonstrated_empathy": 14.8, "depth_of_insight": 15.6, "emotional_reasoning": 14.7, "humanlike": 14.0, "intellectual_grounding": 13.0, "message_tailoring": 13.1, "moralising": 8.0, "pragmatic_ei": 13.7, "reactive": 8.7, "safety_conscious": 12.4, "social_dexterity": 13.4, "subtext_identification": 15.5, "sycophantic": 3.6, "theory_of_mind": 13.6, "validating": 12.9, "warmth": 13.4 }, "public_source_key": "claude-sonnet-4-6", "public_rubric_0_100": 80.0, "public_elo_norm": 1714.1, "public_traits_17": { "demonstrated_empathy": 16.15, "pragmatic_ei": 16.0, "depth_of_insight": 17.29, "social_dexterity": 15.19, "emotional_reasoning": 16.64, "message_tailoring": 14.65, "boundary_setting": 13.77, "safety_conscious": 13.35, "moralising": 6.08, "compliant": 9.88, "challenging": 12.58, "warmth": 13.77, "validating": 13.27, "analytical": 17.46, "reactive": 8.88, "conversational": 11.31, "humanlike": 15.35 } }, "pinchbench": { "best_score": 88.0, "avg_score": 81.1, "runs": 19 }, "arena": { "elo": 1462, "rank": 18, "category": "text" } }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": true, "eq_bench_public": true }, "created": 1771342990, "last_updated": "2026-10-12T09:47:31Z", "speed": { "output_tokens_per_sec": 47.0, "ttft_seconds": 0.9, "source": "artificial_analysis" } }, { "id": "deepseek/deepseek-v4-flash", "canonical_slug": "deepseek/deepseek-v4-flash-20260423", "name": "DeepSeek V4 Flash 0423", "provider": "DeepSeek", "description": "DeepSeek V4 Flash is an efficiency-optimized Mixture-of-Experts model from DeepSeek with 284B total parameters and 13B activated parameters, supporting a 1M-token context window. It is designed for fast inference and...", "context_window": 1048576, "max_output": 943718, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "tokenizer": "DeepSeek", "pricing": { "input_per_m": 0.03, "output_per_m": 1.28, "cache_read_per_m": 0.03, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 0.3425 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": false, "audio_input": false, "web_search": false, "structured_output": true }, "endpoint_stats": { "providers": [ "Alibaba", "AtlasCloud", "Azure", "Baidu", "Cloudflare", "DeepInfra", "DigitalOcean", "GMICloud", "Mancer 2", "OpenInference", "Parasail", "Relace", "SiliconFlow", "StreamLake", "Venice", "Wafer" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "deepseek-ai/DeepSeek-V4-Flash", "public_rubric_0_100": 72.5, "public_elo_norm": 1491.3, "public_traits_17": { "demonstrated_empathy": 14.65, "pragmatic_ei": 14.04, "depth_of_insight": 16.07, "social_dexterity": 13.31, "emotional_reasoning": 15.18, "message_tailoring": 12.73, "boundary_setting": 12.12, "safety_conscious": 11.88, "moralising": 8.5, "compliant": 11.69, "challenging": 11.38, "warmth": 13.88, "validating": 13.58, "analytical": 16.54, "reactive": 10.73, "conversational": 11.54, "humanlike": 13.88 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1777000666, "last_updated": "2026-10-12T09:47:31Z" }, { "id": "deepseek/deepseek-v4-pro", "canonical_slug": "deepseek/deepseek-v4-pro-20260423", "name": "DeepSeek V4 Pro 0423", "provider": "DeepSeek", "description": "DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. It is designed for advanced reasoning, coding,...", "context_window": 1048576, "max_output": 384000, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "tokenizer": "DeepSeek", "pricing": { "input_per_m": 0.9553, "output_per_m": 1.9105, "cache_read_per_m": 0.0796, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 1.1941 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": false, "audio_input": false, "web_search": false, "structured_output": true }, "endpoint_stats": { "providers": [ "Alibaba", "AtlasCloud", "Azure", "Baidu", "Cloudflare", "DeepInfra", "DigitalOcean", "GMICloud", "Novita", "Parasail", "Reka", "SiliconFlow", "StreamLake", "Venice" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "deepseek-ai/DeepSeek-V4-Pro", "public_rubric_0_100": 77.6, "public_elo_norm": 1570.2, "public_traits_17": { "demonstrated_empathy": 15.73, "pragmatic_ei": 15.04, "depth_of_insight": 17.29, "social_dexterity": 14.5, "emotional_reasoning": 16.38, "message_tailoring": 13.92, "boundary_setting": 12.92, "safety_conscious": 12.58, "moralising": 8.5, "compliant": 11.31, "challenging": 11.96, "warmth": 14.38, "validating": 13.69, "analytical": 17.58, "reactive": 10.23, "conversational": 12.04, "humanlike": 15.23 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1777000679, "last_updated": "2026-10-12T09:47:31Z" }, { "id": "google/gemini-3.1-pro-preview-20260219", "canonical_slug": "google/gemini-3.1-pro-preview-20260219", "name": "Gemini 3.1 Pro Preview", "provider": "Google", "description": "Gemini 3.1 Pro Preview is Google\u2019s frontier reasoning model, delivering enhanced software engineering performance, improved agentic reliability, and more efficient token usage across complex workflows. Building on the multimodal foundation...", "context_window": 1048576, "max_output": 65536, "modalities": { "input": [ "audio", "file", "image", "text", "video" ], "output": [ "text" ] }, "tokenizer": "Gemini", "pricing": { "input_per_m": 2.0, "output_per_m": 12.0, "cache_read_per_m": 0.2, "cache_write_per_m": 0.375, "web_search_per_req": 0.014, "blended_per_m": 4.5 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": true, "web_search": true, "structured_output": true }, "endpoint_stats": { "providers": [ "Google", "Google AI Studio" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": { "reasoning": 57.0, "coding": 56, "agentic": 83.5 }, "benchmarks": { "artificial_analysis": { "intelligence_index": 57.0, "note": "Coding index unavailable \u2014 AA API rate limited during collection" }, "eq_bench": { "elo": 1548.7, "humanlike": 6.6, "safety": 6.6, "assertive": 6.6, "social_iq": 6.8, "warm": 7.4, "analytic": 8.9, "insight": 8.5, "empathy": 7.9, "compliant": 5.9, "moralising": 5.1, "pragmatic": 7.1, "v3_score": 68.95, "v3_traits": { "analytical": 16.3, "boundary_setting": 12.0, "challenging": 11.1, "compliant": 10.2, "conversational": 11.5, "correctness": 13.7, "demonstrated_empathy": 14.6, "depth_of_insight": 15.1, "emotional_reasoning": 14.3, "humanlike": 13.0, "intellectual_grounding": 12.5, "message_tailoring": 13.0, "moralising": 9.8, "pragmatic_ei": 13.0, "reactive": 10.6, "safety_conscious": 12.6, "social_dexterity": 12.8, "subtext_identification": 14.3, "sycophantic": 5.8, "theory_of_mind": 12.8, "validating": 14.4, "warmth": 13.8 }, "public_source_key": "gemini-3.1-pro-preview", "public_rubric_0_100": 74.3, "public_elo_norm": 1537.7, "public_traits_17": { "demonstrated_empathy": 15.19, "pragmatic_ei": 14.23, "depth_of_insight": 16.6, "social_dexterity": 13.35, "emotional_reasoning": 15.6, "message_tailoring": 13.38, "boundary_setting": 13.42, "safety_conscious": 12.92, "moralising": 9.62, "compliant": 11.42, "challenging": 12.31, "warmth": 13.92, "validating": 14.0, "analytical": 17.38, "reactive": 11.54, "conversational": 11.15, "humanlike": 14.04 } }, "pinchbench": { "best_score": 86.7, "avg_score": 77.0, "runs": 15 }, "arena": { "elo": 1494, "rank": 3, "category": "text" } }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1771509627, "last_updated": "2026-10-12T09:47:31Z", "speed": { "output_tokens_per_sec": 127.0, "ttft_seconds": 30.66, "source": "artificial_analysis", "note": "TTFT includes reasoning/thinking time" } }, { "id": "google/gemma-4-26b-a4b-it", "canonical_slug": "google/gemma-4-26b-a4b-it-20260403", "name": "Gemma 4 26B A4B ", "provider": "Google", "description": "Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts (MoE) model from Google DeepMind. Despite 25.2B total parameters, only 3.8B activate per token during inference \u2014 delivering near-31B quality at...", "context_window": 262144, "max_output": 235929, "modalities": { "input": [ "image", "text", "video" ], "output": [ "text" ] }, "tokenizer": "Gemma", "pricing": { "input_per_m": 0.09, "output_per_m": 0.3, "cache_read_per_m": 0.05, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 0.1425 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": false, "structured_output": true }, "endpoint_stats": { "providers": [ "Cloudflare", "CoreWeave", "Darkbloom", "DeepInfra", "DekaLLM", "Google", "Io Net", "Makora", "NextBit", "Novita", "Parasail", "SiliconFlow", "Venice" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "google/gemma-4-26B-A4B-it", "public_rubric_0_100": 70.0, "public_elo_norm": 1430.6, "public_traits_17": { "demonstrated_empathy": 14.85, "pragmatic_ei": 14.54, "depth_of_insight": 15.42, "social_dexterity": 13.42, "emotional_reasoning": 14.51, "message_tailoring": 12.96, "boundary_setting": 13.35, "safety_conscious": 12.58, "moralising": 9.38, "compliant": 11.69, "challenging": 11.54, "warmth": 13.0, "validating": 14.08, "analytical": 17.42, "reactive": 10.81, "conversational": 9.62, "humanlike": 12.12 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1775227989, "last_updated": "2026-10-12T09:47:31Z" }, { "id": "google/gemma-4-31b-it", "canonical_slug": "google/gemma-4-31b-it-20260402", "name": "Gemma 4 31B", "provider": "Google", "description": "Gemma 4 31B Instruct is Google DeepMind's 30.7B dense multimodal model supporting text and image input with text output. Features a 256K token context window, configurable thinking/reasoning mode, native function...", "context_window": 262144, "max_output": 16384, "modalities": { "input": [ "image", "text", "video" ], "output": [ "text" ] }, "tokenizer": "Gemma", "pricing": { "input_per_m": 0.09, "output_per_m": 0.34, "cache_read_per_m": 0.05, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 0.1525 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": false, "structured_output": true }, "endpoint_stats": { "providers": [ "Chutes", "CoreWeave", "Crusoe", "DeepInfra", "Friendli", "Io Net", "ModelRun", "Novita", "Parasail", "SambaNova", "SiliconFlow", "Venice" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": { "reasoning": 52.3, "coding": 38.7, "agentic": 75.6 }, "benchmarks": { "artificial_analysis": { "intelligence_index": 39.2, "coding_index": 38.7, "gpqa": 0.857, "hle": 0.227, "scicode": 0.434, "ifbench": 0.756, "terminalbench_hard": 0.364 }, "eq_bench": { "v3_score": 66.1, "v3_traits": { "analytical": 16.2, "boundary_setting": 13.1, "challenging": 12.2, "compliant": 9.3, "conversational": 10.6, "correctness": 12.9, "demonstrated_empathy": 14.0, "depth_of_insight": 14.2, "emotional_reasoning": 13.6, "humanlike": 11.8, "intellectual_grounding": 11.3, "message_tailoring": 12.8, "moralising": 10.7, "pragmatic_ei": 14.0, "reactive": 10.2, "safety_conscious": 11.5, "social_dexterity": 12.9, "subtext_identification": 12.6, "sycophantic": 4.2, "theory_of_mind": 11.7, "validating": 12.5, "warmth": 11.7 }, "public_source_key": "google/gemma-4-31B-it", "public_rubric_0_100": 70.8, "public_elo_norm": 1434.1, "public_traits_17": { "demonstrated_empathy": 14.73, "pragmatic_ei": 15.08, "depth_of_insight": 15.38, "social_dexterity": 13.88, "emotional_reasoning": 14.62, "message_tailoring": 13.65, "boundary_setting": 14.31, "safety_conscious": 12.62, "moralising": 10.19, "compliant": 10.62, "challenging": 12.85, "warmth": 12.0, "validating": 12.38, "analytical": 17.31, "reactive": 10.88, "conversational": 10.46, "humanlike": 12.77 } }, "pinchbench": {}, "arena": { "elo": 1452, "note": "Preliminary" } }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1775148486, "last_updated": "2026-10-12T09:47:31Z", "speed": { "output_tokens_per_sec": 35.9, "ttft_seconds": null, "ttfa_seconds": null } }, { "id": "minimax/minimax-m2.7", "canonical_slug": "minimax/minimax-m2.7-20260318", "name": "MiniMax M2.7", "provider": "MiniMax", "description": "MiniMax-M2.7 is a next-generation large language model designed for autonomous, real-world productivity and continuous improvement. Built to actively participate in its own evolution, M2.7 integrates advanced agentic capabilities through multi-agent...", "context_window": 204800, "max_output": 176947, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "tokenizer": "Other", "pricing": { "input_per_m": 0.21, "output_per_m": 0.84, "cache_read_per_m": 0.042, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 0.3675 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": false, "audio_input": false, "web_search": false, "structured_output": true }, "endpoint_stats": { "providers": [ "AtlasCloud", "GMICloud", "Groq", "Minimax", "Novita" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": { "reasoning": 58.5, "coding": 41.9, "agentic": 87.6 }, "benchmarks": { "artificial_analysis": { "intelligence_index": 49.6, "coding_index": 41.9, "gpqa": 0.874, "hle": 0.281, "scicode": 0.47, "ifbench": 0.757, "terminalbench_hard": 0.394 }, "eq_bench": { "v3_score": 68.75, "v3_traits": { "analytical": 17.8, "boundary_setting": 10.5, "challenging": 10.4, "compliant": 10.8, "conversational": 10.9, "correctness": 14.6, "demonstrated_empathy": 14.3, "depth_of_insight": 15.5, "emotional_reasoning": 14.8, "humanlike": 13.7, "intellectual_grounding": 12.9, "message_tailoring": 11.2, "moralising": 5.4, "pragmatic_ei": 11.2, "reactive": 10.1, "safety_conscious": 12.7, "social_dexterity": 10.6, "subtext_identification": 16.3, "sycophantic": 3.8, "theory_of_mind": 15.1, "validating": 14.1, "warmth": 12.4 } }, "pinchbench": { "best_score": 89.8, "avg_score": 83.2, "runs": 11 }, "arena": { "elo": 1403, "rank": 86, "category": "text" } }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": true }, "created": 1773836697, "last_updated": "2026-10-12T09:47:32Z", "speed": { "output_tokens_per_sec": 41.9, "ttft_seconds": 1.385, "source": "artificial_analysis" } }, { "id": "moonshotai/kimi-k2", "canonical_slug": "moonshotai/kimi-k2", "name": "Kimi K2 0711", "provider": "MoonshotAI", "description": "Kimi K2 Instruct is a large-scale Mixture-of-Experts (MoE) language model developed by Moonshot AI, featuring 1 trillion total parameters with 32 billion active per forward pass. It is optimized for...", "context_window": 131072, "max_output": 98304, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "tokenizer": "Other", "pricing": { "input_per_m": 0.57, "output_per_m": 2.3, "cache_read_per_m": null, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 1.0025 }, "capabilities": { "tool_use": true, "reasoning": false, "vision": false, "audio_input": false, "web_search": false, "structured_output": false }, "endpoint_stats": { "providers": [ "Novita" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "moonshotai/Kimi-K2-Instruct", "public_rubric_0_100": 75.0, "public_elo_norm": 1562.2, "public_traits_17": { "demonstrated_empathy": 15.12, "pragmatic_ei": 14.69, "depth_of_insight": 16.96, "social_dexterity": 13.73, "emotional_reasoning": 15.89, "message_tailoring": 13.15, "boundary_setting": 13.65, "safety_conscious": 13.46, "moralising": 8.42, "compliant": 11.31, "challenging": 11.58, "warmth": 13.81, "validating": 13.58, "analytical": 17.46, "reactive": 10.73, "conversational": 10.54, "humanlike": 14.04 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1752263252, "last_updated": "2026-10-12T09:47:32Z" }, { "id": "moonshotai/kimi-k2.5", "canonical_slug": "moonshotai/kimi-k2.5-0127", "name": "Kimi K2.5", "provider": "MoonshotAI", "description": "Kimi K2.5 is Moonshot AI's native multimodal model, delivering state-of-the-art visual coding capability and a self-directed agent swarm paradigm. Built on Kimi K2 with continued pretraining over approximately 15T mixed...", "context_window": 262144, "max_output": 235929, "modalities": { "input": [ "text", "image" ], "output": [ "text" ] }, "tokenizer": "Other", "pricing": { "input_per_m": 0.5, "output_per_m": 2.5, "cache_read_per_m": 0.15, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 1.0 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": false, "structured_output": true }, "endpoint_stats": { "providers": [ "Amazon Bedrock", "AtlasCloud", "Novita", "SiliconFlow", "Venice" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "moonshotai/Kimi-K2.5", "public_rubric_0_100": 72.5, "public_elo_norm": 1544.7, "public_traits_17": { "demonstrated_empathy": 14.58, "pragmatic_ei": 13.54, "depth_of_insight": 16.78, "social_dexterity": 12.65, "emotional_reasoning": 15.44, "message_tailoring": 12.23, "boundary_setting": 14.38, "safety_conscious": 13.77, "moralising": 11.12, "compliant": 10.04, "challenging": 14.15, "warmth": 12.23, "validating": 12.19, "analytical": 17.77, "reactive": 11.54, "conversational": 9.81, "humanlike": 13.19 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1769487076, "last_updated": "2026-10-12T09:47:32Z" }, { "id": "moonshotai/kimi-k2.6", "canonical_slug": "moonshotai/kimi-k2.6-20260420", "name": "Kimi K2.6", "provider": "MoonshotAI", "description": "Kimi K2.6 is Moonshot AI's next-generation multimodal model, designed for long-horizon coding, coding-driven UI/UX generation, and multi-agent orchestration. It handles complex end-to-end coding tasks across Python, Rust, and Go, and...", "context_window": 262144, "max_output": 235929, "modalities": { "input": [ "text", "image" ], "output": [ "text" ] }, "tokenizer": "Other", "pricing": { "input_per_m": 0.65, "output_per_m": 3.41, "cache_read_per_m": 0.15, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 1.34 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": false, "structured_output": true }, "endpoint_stats": { "providers": [ "AtlasCloud", "Baidu", "Chutes", "Cloudflare", "CoreWeave", "Crusoe", "DeepInfra", "DigitalOcean", "GMICloud", "Inceptron", "Moonshot AI", "Novita", "Parasail", "Phala", "SiliconFlow", "StreamLake", "Venice" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "moonshotai/Kimi-K2.6", "public_rubric_0_100": 76.6, "public_elo_norm": 1561.2, "public_traits_17": { "demonstrated_empathy": 15.08, "pragmatic_ei": 13.46, "depth_of_insight": 17.64, "social_dexterity": 12.65, "emotional_reasoning": 16.67, "message_tailoring": 12.5, "boundary_setting": 14.69, "safety_conscious": 13.5, "moralising": 12.0, "compliant": 10.58, "challenging": 13.88, "warmth": 12.46, "validating": 12.31, "analytical": 18.04, "reactive": 11.96, "conversational": 9.69, "humanlike": 14.08 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1776699402, "last_updated": "2026-10-12T09:47:32Z" }, { "id": "nousresearch/hermes-4-405b", "canonical_slug": "nousresearch/hermes-4-405b", "name": "Hermes 4 405B", "provider": "Nous", "description": "Hermes 4 is a large-scale reasoning model built on Meta-Llama-3.1-405B and released by Nous Research. It introduces a hybrid reasoning mode, where the model can choose to deliberate internally with...", "context_window": 131072, "max_output": 117964, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "tokenizer": "Other", "pricing": { "input_per_m": 1.0, "output_per_m": 3.0, "cache_read_per_m": null, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 1.5 }, "capabilities": { "tool_use": false, "reasoning": true, "vision": false, "audio_input": false, "web_search": false, "structured_output": true }, "endpoint_stats": { "providers": [ "Nebius" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "NousResearch/Hermes-4-405B", "public_rubric_0_100": 65.2, "public_elo_norm": 1361.4, "public_traits_17": { "demonstrated_empathy": 13.88, "pragmatic_ei": 13.54, "depth_of_insight": 14.16, "social_dexterity": 12.5, "emotional_reasoning": 13.69, "message_tailoring": 12.31, "boundary_setting": 12.38, "safety_conscious": 11.46, "moralising": 9.0, "compliant": 12.0, "challenging": 10.73, "warmth": 13.0, "validating": 13.85, "analytical": 16.42, "reactive": 10.88, "conversational": 10.23, "humanlike": 11.96 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1756235463, "last_updated": "2026-10-12T09:47:32Z" }, { "id": "openai/gpt-5.1", "canonical_slug": "openai/gpt-5.1-20251113", "name": "GPT-5.1", "provider": "OpenAI", "description": "GPT-5.1 is the latest frontier-grade model in the GPT-5 series, offering stronger general-purpose reasoning, improved instruction adherence, and a more natural conversational style compared to GPT-5. It uses adaptive reasoning...", "context_window": 400000, "max_output": 128000, "modalities": { "input": [ "image", "text", "file" ], "output": [ "text" ] }, "tokenizer": "GPT", "pricing": { "input_per_m": 1.25, "output_per_m": 10.0, "cache_read_per_m": 0.125, "cache_write_per_m": null, "web_search_per_req": 0.01, "blended_per_m": 3.4375 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": true, "structured_output": true }, "endpoint_stats": { "providers": [ "Azure", "OpenAI" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "gpt-5.1-2025-11-13", "public_rubric_0_100": 78.1, "public_elo_norm": 1551.0, "public_traits_17": { "demonstrated_empathy": 15.35, "pragmatic_ei": 15.85, "depth_of_insight": 16.91, "social_dexterity": 14.69, "emotional_reasoning": 16.29, "message_tailoring": 13.92, "boundary_setting": 14.42, "safety_conscious": 13.19, "moralising": 7.15, "compliant": 11.58, "challenging": 11.69, "warmth": 13.0, "validating": 14.31, "analytical": 17.46, "reactive": 9.15, "conversational": 10.5, "humanlike": 13.92 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1763060305, "last_updated": "2026-10-12T09:47:32Z" }, { "id": "openai/gpt-5.2", "canonical_slug": "openai/gpt-5.2-20251211", "name": "GPT-5.2", "provider": "OpenAI", "description": "GPT-5.2 is the latest frontier-grade model in the GPT-5 series, offering stronger agentic and long context perfomance compared to GPT-5.1. It uses adaptive reasoning to allocate computation dynamically, responding quickly...", "context_window": 400000, "max_output": 128000, "modalities": { "input": [ "file", "image", "text" ], "output": [ "text" ] }, "tokenizer": "GPT", "pricing": { "input_per_m": 1.75, "output_per_m": 14.0, "cache_read_per_m": 0.175, "cache_write_per_m": null, "web_search_per_req": 0.01, "blended_per_m": 4.8125 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": true, "structured_output": true }, "endpoint_stats": { "providers": [ "Azure", "OpenAI" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "gpt-5.2", "public_rubric_0_100": 80.4, "public_elo_norm": 1558.8, "public_traits_17": { "demonstrated_empathy": 15.35, "pragmatic_ei": 16.15, "depth_of_insight": 17.4, "social_dexterity": 14.81, "emotional_reasoning": 16.64, "message_tailoring": 14.04, "boundary_setting": 15.54, "safety_conscious": 14.5, "moralising": 7.31, "compliant": 10.54, "challenging": 13.04, "warmth": 12.31, "validating": 13.65, "analytical": 18.0, "reactive": 8.88, "conversational": 10.19, "humanlike": 13.54 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1765389775, "last_updated": "2026-10-12T09:47:32Z" }, { "id": "openai/gpt-5.3-chat", "canonical_slug": "openai/gpt-5.3-chat-20260303", "name": "GPT-5.3 Chat", "provider": "OpenAI", "description": "GPT-5.3 Chat is an update to ChatGPT's most-used model that makes everyday conversations smoother, more useful, and more directly helpful. It delivers more accurate answers with better contextualization and significantly...", "context_window": 128000, "max_output": 16384, "modalities": { "input": [ "text", "image", "file" ], "output": [ "text" ] }, "tokenizer": "GPT", "pricing": { "input_per_m": 1.75, "output_per_m": 14.0, "cache_read_per_m": 0.175, "cache_write_per_m": null, "web_search_per_req": 0.01, "blended_per_m": 4.8125 }, "capabilities": { "tool_use": true, "reasoning": false, "vision": true, "audio_input": false, "web_search": true, "structured_output": true }, "endpoint_stats": { "providers": [ "OpenAI" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "gpt-5.3-chat", "public_rubric_0_100": 70.7, "public_elo_norm": 1393.0, "public_traits_17": { "demonstrated_empathy": 12.88, "pragmatic_ei": 13.35, "depth_of_insight": 15.62, "social_dexterity": 12.42, "emotional_reasoning": 14.69, "message_tailoring": 11.65, "boundary_setting": 11.77, "safety_conscious": 11.85, "moralising": 5.65, "compliant": 11.04, "challenging": 8.85, "warmth": 11.58, "validating": 12.96, "analytical": 16.5, "reactive": 7.73, "conversational": 10.19, "humanlike": 11.35 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1772564061, "last_updated": "2026-08-10T10:36:16Z" }, { "id": "openai/gpt-5.4", "canonical_slug": "openai/gpt-5.4-20260305", "name": "GPT-5.4", "provider": "OpenAI", "description": "GPT-5.4 is OpenAI\u2019s latest frontier model, unifying the Codex and GPT lines into a single system. It features a 1M+ token context window (922K input, 128K output) with support for...", "context_window": 1050000, "max_output": 128000, "modalities": { "input": [ "text", "image", "file" ], "output": [ "text" ] }, "tokenizer": "GPT", "pricing": { "input_per_m": 2.5, "output_per_m": 15.0, "cache_read_per_m": 0.25, "cache_write_per_m": null, "web_search_per_req": 0.01, "blended_per_m": 5.625 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": true, "structured_output": true }, "endpoint_stats": { "providers": [ "Amazon Bedrock", "Azure", "OpenAI" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": { "reasoning": 57.0, "coding": 57, "agentic": 87.6 }, "benchmarks": { "artificial_analysis": { "intelligence_index": 57.0, "note": "Coding index unavailable \u2014 AA API rate limited during collection" }, "eq_bench": { "elo": 1687.5, "humanlike": 6.3, "safety": 7.5, "assertive": 7.5, "social_iq": 7.7, "warm": 6.4, "analytic": 9.4, "insight": 9.4, "empathy": 8.2, "compliant": 5.0, "moralising": 3.9, "pragmatic": 8.7, "v3_score": 73.2, "v3_traits": { "analytical": 16.8, "boundary_setting": 14.3, "challenging": 11.8, "compliant": 7.3, "conversational": 9.5, "correctness": 14.8, "demonstrated_empathy": 14.4, "depth_of_insight": 15.8, "emotional_reasoning": 15.1, "humanlike": 12.3, "intellectual_grounding": 13.6, "message_tailoring": 13.0, "moralising": 9.6, "pragmatic_ei": 14.8, "reactive": 7.3, "safety_conscious": 13.5, "social_dexterity": 13.6, "subtext_identification": 15.6, "sycophantic": 3.2, "theory_of_mind": 14.7, "validating": 13.1, "warmth": 11.8 }, "public_source_key": "gpt-5.4", "public_rubric_0_100": 82.4, "public_elo_norm": 1562.9, "public_traits_17": { "demonstrated_empathy": 15.77, "pragmatic_ei": 16.62, "depth_of_insight": 17.84, "social_dexterity": 15.19, "emotional_reasoning": 17.18, "message_tailoring": 14.5, "boundary_setting": 15.5, "safety_conscious": 14.27, "moralising": 7.58, "compliant": 10.19, "challenging": 12.88, "warmth": 12.15, "validating": 13.65, "analytical": 17.96, "reactive": 8.65, "conversational": 10.46, "humanlike": 14.31 } }, "pinchbench": { "best_score": 90.5, "avg_score": 81.7, "runs": 17 }, "arena": { "elo": 1484, "rank": 6, "category": "text" } }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1772734352, "last_updated": "2026-10-12T09:47:32Z", "speed": { "output_tokens_per_sec": 74.0, "ttft_seconds": 152.2, "source": "artificial_analysis", "note": "TTFT includes reasoning/thinking time" } }, { "id": "openai/gpt-5.4-mini", "canonical_slug": "openai/gpt-5.4-mini-20260317", "name": "GPT-5.4 Mini", "provider": "OpenAI", "description": "GPT-5.4 mini brings the core capabilities of GPT-5.4 to a faster, more efficient model optimized for high-throughput workloads. It supports text and image inputs with strong performance across reasoning, coding,...", "context_window": 400000, "max_output": 128000, "modalities": { "input": [ "file", "image", "text" ], "output": [ "text" ] }, "tokenizer": "GPT", "pricing": { "input_per_m": 0.75, "output_per_m": 4.5, "cache_read_per_m": 0.075, "cache_write_per_m": null, "web_search_per_req": 0.01, "blended_per_m": 1.6875 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": true, "structured_output": true }, "endpoint_stats": { "providers": [ "Azure", "OpenAI" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": { "reasoning": 48, "coding": 51, "agentic": 56 }, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "v3_score": 68.65, "v3_traits": { "analytical": 18.0, "boundary_setting": 15.5, "challenging": 12.0, "compliant": 8.2, "conversational": 8.7, "correctness": 14.5, "demonstrated_empathy": 13.4, "depth_of_insight": 15.4, "emotional_reasoning": 14.9, "humanlike": 11.4, "intellectual_grounding": 12.6, "message_tailoring": 10.7, "moralising": 6.0, "pragmatic_ei": 13.1, "reactive": 7.7, "safety_conscious": 15.2, "social_dexterity": 11.2, "subtext_identification": 16.0, "sycophantic": 2.7, "theory_of_mind": 14.2, "validating": 10.9, "warmth": 8.6 } }, "pinchbench": {}, "arena": { "elo": 1455 } }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false }, "created": 1773748178, "last_updated": "2026-10-12T09:47:32Z", "notes": "Agentic score from AA Agentic Index (not PinchBench-based)", "speed": { "output_tokens_per_sec": 186, "source": "artificial_analysis" } }, { "id": "openai/gpt-5.5", "canonical_slug": "openai/gpt-5.5-20260423", "name": "GPT-5.5", "provider": "OpenAI", "description": "GPT-5.5 is OpenAI\u2019s frontier model designed for complex professional workloads, building on GPT-5.4 with stronger reasoning, higher reliability, and improved token efficiency on hard tasks. It features a 1M+ token...", "context_window": 1050000, "max_output": 128000, "modalities": { "input": [ "file", "image", "text" ], "output": [ "text" ] }, "tokenizer": "GPT", "pricing": { "input_per_m": 5.0, "output_per_m": 30.0, "cache_read_per_m": 0.5, "cache_write_per_m": null, "web_search_per_req": 0.01, "blended_per_m": 11.25 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": true, "structured_output": true }, "endpoint_stats": { "providers": [ "Amazon Bedrock", "Azure", "OpenAI" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "gpt-5.5", "public_rubric_0_100": 82.4, "public_elo_norm": 1576.9, "public_traits_17": { "demonstrated_empathy": 15.88, "pragmatic_ei": 16.5, "depth_of_insight": 17.67, "social_dexterity": 15.23, "emotional_reasoning": 17.07, "message_tailoring": 14.46, "boundary_setting": 15.77, "safety_conscious": 14.38, "moralising": 8.46, "compliant": 9.65, "challenging": 13.38, "warmth": 13.23, "validating": 13.58, "analytical": 17.69, "reactive": 8.62, "conversational": 10.42, "humanlike": 13.69 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1777051893, "last_updated": "2026-10-12T09:47:32Z" }, { "id": "qwen/qwen3.5-397b-a17b", "canonical_slug": "qwen/qwen3.5-397b-a17b-20260216", "name": "Qwen3.5 397B A17B", "provider": "Qwen", "description": "The Qwen3.5 series 397B-A17B native vision-language model is built on a hybrid architecture that integrates a linear attention mechanism with a sparse mixture-of-experts model, achieving higher inference efficiency. It delivers...", "context_window": 262144, "max_output": 235929, "modalities": { "input": [ "text", "image", "video" ], "output": [ "text" ] }, "tokenizer": "Qwen3", "pricing": { "input_per_m": 0.55, "output_per_m": 3.5, "cache_read_per_m": 0.225, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 1.2875 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": false, "structured_output": true }, "endpoint_stats": { "providers": [ "Alibaba", "AtlasCloud", "DeepInfra", "DigitalOcean", "GMICloud", "Novita", "Parasail", "Phala", "StreamLake", "Venice" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "Qwen/Qwen3.5-397B-A17B", "public_rubric_0_100": 71.3, "public_elo_norm": 1413.1, "public_traits_17": { "demonstrated_empathy": 14.35, "pragmatic_ei": 14.0, "depth_of_insight": 15.87, "social_dexterity": 12.69, "emotional_reasoning": 15.0, "message_tailoring": 12.23, "boundary_setting": 13.65, "safety_conscious": 12.88, "moralising": 9.85, "compliant": 12.12, "challenging": 11.23, "warmth": 13.0, "validating": 13.58, "analytical": 16.81, "reactive": 10.92, "conversational": 9.42, "humanlike": 12.27 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1771223018, "last_updated": "2026-10-12T09:47:32Z" }, { "id": "qwen/qwen3.6-plus:free", "canonical_slug": "qwen/qwen3.6-plus-04-02", "name": "Qwen3.6 Plus", "provider": "Qwen", "description": "Qwen 3.6 Plus builds on a hybrid architecture that combines efficient linear attention with sparse mixture-of-experts routing, enabling strong scalability and high-performance inference. Compared to the 3.5 series, it delivers...", "context_window": 1000000, "max_output": 65536, "modalities": { "input": [ "text", "image", "video" ], "output": [ "text" ] }, "tokenizer": "Qwen3", "pricing": { "input_per_m": 0.325, "output_per_m": 1.95, "cache_read_per_m": null, "cache_write_per_m": 0.4062, "web_search_per_req": null, "blended_per_m": 0.7312 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": false, "structured_output": true }, "endpoint_stats": { "providers": [ "Alibaba" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": { "reasoning": 57.0, "coding": 41.3, "agentic": 87.1 }, "benchmarks": { "artificial_analysis": { "intelligence_index": 45.0, "coding_index": 41.3, "gpqa": 0.893, "hle": 0.273, "scicode": 0.42, "ifbench": 0.788, "terminalbench_hard": 0.409, "note": "Data from Qwen3.5-397B (predecessor)" }, "eq_bench": {}, "pinchbench": { "best_score": 88.6, "avg_score": 84.0, "runs": 5 }, "arena": { "elo": 1449, "rank": 30, "category": "text", "note": "Score from Qwen3.5-397B (predecessor)" } }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false }, "created": 1775133557, "last_updated": "2026-10-12T09:47:32Z", "notes": "EQ-Bench scores removed 2026-07-26: they were inherited from predecessor Qwen3.5-397B-A17B, which is now tracked as its own row (qwen/qwen3.5-397b-a17b). Qwen3.6 Plus has not been benchmarked on EQ-Bench. PinchBench is from Qwen3.6 Plus directly. The :free OpenRouter tier has been retired; pricing shown is the paid qwen/qwen3.6-plus endpoint.", "speed": { "output_tokens_per_sec": 44.0, "ttft_seconds": 1.59, "source": "openrouter" } }, { "id": "x-ai/grok-4.20-20260309", "canonical_slug": "x-ai/grok-4.20-20260309", "name": "Grok 4.20", "provider": "SpaceXAI", "description": "Grok 4.20 is a reasoning model from SpaceXAI with industry-leading speed and agentic tool calling capabilities. It combines the lowest hallucination rate on the market with strict prompt adherance, delivering...", "context_window": 2000000, "max_output": 1800000, "modalities": { "input": [ "text", "image", "file" ], "output": [ "text" ] }, "tokenizer": "Grok", "pricing": { "input_per_m": 1.25, "output_per_m": 2.5, "cache_read_per_m": 0.2, "cache_write_per_m": null, "web_search_per_req": 0.005, "blended_per_m": 1.5625 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": true, "audio_input": false, "web_search": true, "structured_output": true }, "endpoint_stats": { "providers": [ "xAI" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": { "reasoning": 48.0, "coding": 42, "agentic": 78.9 }, "benchmarks": { "artificial_analysis": { "intelligence_index": 48.0, "note": "Coding index unavailable \u2014 AA API rate limited during collection" }, "eq_bench": { "elo": 856.4, "v3_score": 68.55, "v3_traits": { "analytical": 16.0, "boundary_setting": 11.7, "challenging": 10.6, "compliant": 10.8, "conversational": 10.0, "correctness": 13.5, "demonstrated_empathy": 13.9, "depth_of_insight": 15.3, "emotional_reasoning": 14.3, "humanlike": 12.4, "intellectual_grounding": 13.5, "message_tailoring": 11.5, "moralising": 11.4, "pragmatic_ei": 12.2, "reactive": 9.9, "safety_conscious": 12.2, "social_dexterity": 11.9, "subtext_identification": 15.8, "sycophantic": 4.9, "theory_of_mind": 14.2, "validating": 13.5, "warmth": 13.5 }, "public_source_key": "grok-4.20-beta", "public_rubric_0_100": 55.8, "public_elo_norm": 851.5, "public_traits_17": { "demonstrated_empathy": 7.38, "pragmatic_ei": 7.54, "depth_of_insight": 12.49, "social_dexterity": 6.85, "emotional_reasoning": 11.69, "message_tailoring": 6.31, "boundary_setting": 15.31, "safety_conscious": 14.38, "moralising": 12.54, "compliant": 6.96, "challenging": 11.85, "warmth": 6.77, "validating": 7.73, "analytical": 10.88, "reactive": 13.19, "conversational": 5.77, "humanlike": 7.65 } }, "pinchbench": { "best_score": 82.4, "avg_score": 71.8, "runs": 18, "note": "Score from grok-4.1-fast (closest available)" }, "arena": { "elo": 1491, "rank": 4, "category": "text" } }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1774979019, "last_updated": "2026-10-12T09:47:32Z", "notes": "EQ-Bench score is notably low (856) despite strong Arena ranking \u2014 suggests optimized for capability over emotional attunement.", "speed": { "output_tokens_per_sec": 271.0, "ttft_seconds": 10.72, "source": "artificial_analysis" } }, { "id": "stepfun/step-3.5-flash", "canonical_slug": "stepfun/step-3.5-flash", "name": "Step 3.5 Flash", "provider": "StepFun", "description": "Step 3.5 Flash is StepFun's most capable open-source foundation model. Built on a sparse Mixture of Experts (MoE) architecture, it selectively activates only 11B of its 196B parameters per token....", "context_window": 262144, "max_output": 65536, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "tokenizer": "Other", "pricing": { "input_per_m": 0.1, "output_per_m": 0.3, "cache_read_per_m": null, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 0.15 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": false, "audio_input": false, "web_search": false, "structured_output": false }, "endpoint_stats": { "providers": [ "SiliconFlow" ], "uptime_24h": 99.39, "latency_ms": null, "throughput_tps": null }, "scores": { "reasoning": 50.0, "coding": 31.6, "agentic": 82.5 }, "benchmarks": { "artificial_analysis": { "intelligence_index": 37.8, "coding_index": 31.6, "gpqa": 0.831, "hle": 0.191, "scicode": 0.404, "ifbench": 0.646, "terminalbench_hard": 0.273 }, "eq_bench": { "v3_score": 69.25 }, "pinchbench": { "best_score": 85.3, "avg_score": 76.9, "runs": 18 }, "arena": { "elo": 1392, "rank": 102, "category": "text" } }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": true }, "created": 1769728337, "last_updated": "2026-10-12T09:47:32Z", "speed": { "output_tokens_per_sec": 85.7, "ttft_seconds": 1.271, "source": "artificial_analysis" } }, { "id": "xiaomi/mimo-v2-pro", "name": "MiMo-V2-Pro", "provider": "Xiaomi", "description": "Xiaomi's flagship foundation model with 1T+ parameters and 1M context length, deeply optimized for agentic scenarios.", "context_window": 1048576, "max_output": 131072, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "pricing": { "input_per_m": 1.0, "output_per_m": 3.0, "blended_per_m": 1.5, "cache_read_per_m": 0.2, "cache_write_per_m": null }, "capabilities": { "tool_use": true, "reasoning": true, "vision": false, "web_search": false, "structured_output": true }, "speed": { "output_tokens_per_sec": 35.0, "ttft_seconds": 2.07, "source": "openrouter" }, "scores": { "reasoning": 58.2, "coding": 41.4, "agentic": 82.9 }, "benchmarks": { "artificial_analysis": { "intelligence_index": 49.2, "coding_index": 41.4, "gpqa": 0.87, "hle": 0.283, "scicode": 0.425, "ifbench": 0.688, "terminalbench_hard": 0.409 }, "pinchbench": { "best_score": 83.95, "avg_score": 80.7, "runs": 15 }, "arena": { "elo": 1444, "rank": 36, "category": "text" }, "eq_bench": { "v3_score": 70.55, "v3_traits": { "analytical": 18.1, "boundary_setting": 12.1, "challenging": 11.4, "compliant": 9.2, "conversational": 11.4, "correctness": 13.9, "demonstrated_empathy": 14.8, "depth_of_insight": 15.8, "emotional_reasoning": 14.9, "humanlike": 15.1, "intellectual_grounding": 12.6, "message_tailoring": 12.7, "moralising": 5.7, "pragmatic_ei": 12.8, "reactive": 10.0, "safety_conscious": 13.6, "social_dexterity": 12.0, "subtext_identification": 15.7, "sycophantic": 3.1, "theory_of_mind": 13.9, "validating": 13.2, "warmth": 12.2 } } }, "endpoint_stats": { "providers": [ "Xiaomi" ] }, "last_updated": "2026-04-07", "notes": "Delisted from OpenRouter. Pricing and availability are frozen as of the last successful fetch; successor is xiaomi/mimo-v2.5-pro, which has not been benchmarked here.", "delisted": true }, { "id": "z-ai/glm-4.7", "canonical_slug": "z-ai/glm-4.7-20251222", "name": "GLM 4.7", "provider": "Z.ai", "description": "GLM-4.7 is Z.ai\u2019s latest flagship model, featuring upgrades in two key areas: enhanced programming capabilities and more stable multi-step reasoning/execution. It demonstrates significant improvements in executing complex agent tasks while...", "context_window": 204800, "max_output": 131072, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "tokenizer": "Other", "pricing": { "input_per_m": 0.6, "output_per_m": 2.2, "cache_read_per_m": 0.11, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 1.0 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": false, "audio_input": false, "web_search": false, "structured_output": true }, "endpoint_stats": { "providers": [ "DeepInfra", "Google", "Mancer 2", "Novita", "Venice", "Z.AI" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "zai-org/GLM-4.7", "public_rubric_0_100": 68.9, "public_elo_norm": 1428.3, "public_traits_17": { "demonstrated_empathy": 13.96, "pragmatic_ei": 13.85, "depth_of_insight": 15.49, "social_dexterity": 12.54, "emotional_reasoning": 14.47, "message_tailoring": 12.23, "boundary_setting": 13.15, "safety_conscious": 12.58, "moralising": 9.08, "compliant": 11.46, "challenging": 11.38, "warmth": 11.88, "validating": 13.15, "analytical": 17.08, "reactive": 11.42, "conversational": 10.81, "humanlike": 13.27 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1766378014, "last_updated": "2026-10-12T09:47:32Z" }, { "id": "z-ai/glm-4.7-flash", "canonical_slug": "z-ai/glm-4.7-flash-20260119", "name": "GLM 4.7 Flash", "provider": "Z.ai", "description": "As a 30B-class SOTA model, GLM-4.7-Flash offers a new option that balances performance and efficiency. It is further optimized for agentic coding use cases, strengthening coding capabilities, long-horizon task planning,...", "context_window": 200000, "max_output": 117964, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "tokenizer": "Other", "pricing": { "input_per_m": 0.0605, "output_per_m": 0.4, "cache_read_per_m": null, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 0.1454 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": false, "audio_input": false, "web_search": false, "structured_output": true }, "endpoint_stats": { "providers": [ "Cloudflare", "Novita", "Venice" ], "uptime_24h": 99.86, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "zai-org/GLM-4.7-Flash", "public_rubric_0_100": 52.5, "public_elo_norm": 1083.0, "public_traits_17": { "demonstrated_empathy": 10.72, "pragmatic_ei": 9.96, "depth_of_insight": 12.55, "social_dexterity": 8.72, "emotional_reasoning": 11.5, "message_tailoring": 8.84, "boundary_setting": 12.52, "safety_conscious": 10.12, "moralising": 11.44, "compliant": 11.2, "challenging": 12.04, "warmth": 9.36, "validating": 11.2, "analytical": 15.88, "reactive": 13.0, "conversational": 8.0, "humanlike": 9.8 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1768833913, "last_updated": "2026-10-12T09:47:32Z" }, { "id": "z-ai/glm-5", "canonical_slug": "z-ai/glm-5-20260211", "name": "GLM 5", "provider": "Z.ai", "description": "GLM-5 is Z.ai\u2019s flagship open-source foundation model engineered for complex systems design and long-horizon agent workflows. Built for expert developers, it delivers production-grade performance on large-scale programming tasks, rivaling leading...", "context_window": 204800, "max_output": 128000, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "tokenizer": "Other", "pricing": { "input_per_m": 0.6, "output_per_m": 1.92, "cache_read_per_m": 0.12, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 0.93 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": false, "audio_input": false, "web_search": false, "structured_output": true }, "endpoint_stats": { "providers": [ "Amazon Bedrock", "Baidu", "GMICloud", "Novita", "SiliconFlow", "StreamLake", "Venice", "Z.AI" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "zai-org/GLM-5", "public_rubric_0_100": 75.0, "public_elo_norm": 1526.0, "public_traits_17": { "demonstrated_empathy": 15.15, "pragmatic_ei": 14.62, "depth_of_insight": 16.71, "social_dexterity": 14.04, "emotional_reasoning": 15.69, "message_tailoring": 13.73, "boundary_setting": 12.77, "safety_conscious": 12.15, "moralising": 8.12, "compliant": 10.69, "challenging": 12.54, "warmth": 13.85, "validating": 13.54, "analytical": 16.92, "reactive": 9.73, "conversational": 12.27, "humanlike": 14.77 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1770829182, "last_updated": "2026-10-12T09:47:32Z" }, { "id": "z-ai/glm-5-turbo", "canonical_slug": "z-ai/glm-5-turbo-20260315", "name": "GLM 5 Turbo", "provider": "Z.ai", "description": "GLM-5 Turbo is a new model from Z.ai designed for fast inference and strong performance in agent-driven environments such as OpenClaw scenarios. It is deeply optimized for real-world agent workflows...", "context_window": 202752, "max_output": 131072, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "tokenizer": "Other", "pricing": { "input_per_m": 1.2, "output_per_m": 4.0, "cache_read_per_m": 0.24, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 1.9 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": false, "audio_input": false, "web_search": false, "structured_output": true }, "endpoint_stats": { "providers": [ "Z.AI" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": { "reasoning": 55.8, "coding": 36.8, "agentic": 84.9 }, "benchmarks": { "artificial_analysis": { "intelligence_index": 46.8, "coding_index": 36.8, "gpqa": 0.847, "hle": 0.254, "scicode": 0.436, "ifbench": 0.732, "terminalbench_hard": 0.333 }, "eq_bench": { "elo": 1631.9, "humanlike": 7.2, "safety": 5.9, "assertive": 6.4, "social_iq": 7.2, "warm": 7.1, "analytic": 8.5, "insight": 8.5, "empathy": 7.8, "compliant": 5.3, "moralising": 4.1, "pragmatic": 7.3, "v3_score": 67.7, "v3_traits": { "analytical": 16.3, "boundary_setting": 11.7, "challenging": 12.6, "compliant": 7.5, "conversational": 10.8, "correctness": 13.9, "demonstrated_empathy": 13.7, "depth_of_insight": 15.2, "emotional_reasoning": 14.2, "humanlike": 13.4, "intellectual_grounding": 12.4, "message_tailoring": 12.0, "moralising": 9.2, "pragmatic_ei": 12.6, "reactive": 9.8, "safety_conscious": 12.0, "social_dexterity": 11.8, "subtext_identification": 14.9, "sycophantic": 3.2, "theory_of_mind": 12.9, "validating": 11.2, "warmth": 11.2 } }, "pinchbench": { "best_score": 86.5, "avg_score": 81.6, "runs": 11 }, "arena": { "elo": 1456, "rank": 21, "category": "text" } }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": true }, "created": 1773583573, "last_updated": "2026-10-12T09:47:32Z", "speed": { "output_tokens_per_sec": 42.0, "ttft_seconds": 1.77, "source": "openrouter" } }, { "id": "z-ai/glm-5.1", "canonical_slug": "z-ai/glm-5.1-20260406", "name": "GLM 5.1", "provider": "Z.ai", "description": "GLM-5.1 delivers a major leap in coding capability, with particularly significant gains in handling long-horizon tasks. Unlike previous models built around minute-level interactions, GLM-5.1 can work independently and continuously on...", "context_window": 204800, "max_output": 128000, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "tokenizer": "Other", "pricing": { "input_per_m": 0.966, "output_per_m": 3.036, "cache_read_per_m": 0.1794, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 1.4835 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": false, "audio_input": false, "web_search": false, "structured_output": true }, "endpoint_stats": { "providers": [ "Alibaba", "AtlasCloud", "Baidu", "Chutes", "Friendli", "GMICloud", "Nebius", "Novita", "Phala", "SiliconFlow", "StreamLake", "Venice", "Z.AI" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "zai-org/GLM-5.1", "public_rubric_0_100": 75.2, "public_elo_norm": 1566.5, "public_traits_17": { "demonstrated_empathy": 14.42, "pragmatic_ei": 14.0, "depth_of_insight": 17.04, "social_dexterity": 13.23, "emotional_reasoning": 16.09, "message_tailoring": 12.5, "boundary_setting": 13.27, "safety_conscious": 12.88, "moralising": 9.92, "compliant": 9.54, "challenging": 13.58, "warmth": 12.62, "validating": 12.35, "analytical": 17.04, "reactive": 10.46, "conversational": 11.38, "humanlike": 14.69 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1775578025, "last_updated": "2026-10-12T09:47:32Z" }, { "id": "z-ai/glm-5.2", "canonical_slug": "z-ai/glm-5.2-20260616", "name": "GLM 5.2", "provider": "Z.ai", "description": "GLM 5.2 is a large-scale reasoning model from Z.ai. It supports text input and output with a 1M-token context window, and is suited for long-horizon agent workflows, project-level software engineering,...", "context_window": 1048576, "max_output": 943718, "modalities": { "input": [ "text" ], "output": [ "text" ] }, "tokenizer": "Other", "pricing": { "input_per_m": 0.06, "output_per_m": 7.0, "cache_read_per_m": 0.059, "cache_write_per_m": null, "web_search_per_req": null, "blended_per_m": 1.795 }, "capabilities": { "tool_use": true, "reasoning": true, "vision": false, "audio_input": false, "web_search": false, "structured_output": true }, "endpoint_stats": { "providers": [ "Alibaba", "AtlasCloud", "Baidu", "BaseTen", "Cloudflare", "CoreWeave", "Decart", "DeepInfra", "DigitalOcean", "Friendli", "GMICloud", "Inceptron", "InferenceNet", "Mistral", "Morph", "Nebius", "Novita", "Parasail", "Phala", "SiliconFlow", "StreamLake", "Together", "Venice", "Wafer", "Z.AI" ], "uptime_24h": 100, "latency_ms": null, "throughput_tps": null }, "scores": {}, "benchmarks": { "artificial_analysis": {}, "eq_bench": { "public_source_key": "zai-org/GLM-5.2", "public_rubric_0_100": 78.7, "public_elo_norm": 1575.1, "public_traits_17": { "demonstrated_empathy": 15.73, "pragmatic_ei": 15.54, "depth_of_insight": 17.29, "social_dexterity": 14.73, "emotional_reasoning": 16.36, "message_tailoring": 14.31, "boundary_setting": 14.04, "safety_conscious": 12.85, "moralising": 8.54, "compliant": 10.15, "challenging": 13.69, "warmth": 13.23, "validating": 13.12, "analytical": 17.54, "reactive": 10.31, "conversational": 12.12, "humanlike": 15.42 } }, "pinchbench": {} }, "sources": { "openrouter": true, "artificial_analysis": false, "eq_bench": false, "pinchbench": false, "eq_bench_public": true }, "created": 1781631930, "last_updated": "2026-10-12T09:47:32Z" } ] }