{"baseline":"Three RTX 5060 Ti cards, 16 GB each. A 16 GB run used one card. A 48 GB run used the whole server.","conditions":"pass@1, no tools, one sample","tests":[{"slug":"cursorbench-4","name":"CursorBench 4.0","group":"Software engineering","unit":"percent","higher":true,"measures":"Longer coding tasks"},{"slug":"deepswe-1-1","name":"DeepSWE v1.1","group":"Software engineering","unit":"percent","higher":true,"measures":"Repository issues"},{"slug":"eebench","name":"EEBench","group":"Electrical engineering","unit":"percent","higher":true,"measures":"Electrical engineering tasks"},{"slug":"aa-briefcase-1-1","name":"AA-Briefcase v1.1","group":"Multi-hour office work","unit":"points","higher":true,"measures":"Long office tasks, points"},{"slug":"terminal-bench-4","name":"Terminal-Bench 4.0","group":"Multi-hour terminal work","unit":"percent","higher":true,"measures":"Long terminal tasks"},{"slug":"harvey-legal","name":"Harvey Legal Agent","group":"Legal work","unit":"percent","higher":true,"measures":"Legal agent tasks"},{"slug":"healthbench-pro","name":"HealthBench Pro","group":"Clinical reasoning","unit":"percent","higher":true,"measures":"Clinical reasoning"},{"slug":"price-in","name":"$ per million","group":"Input token price","unit":"usd","higher":false,"fileable":false,"measures":"Published input price"},{"slug":"price-out","name":"$ per million","group":"Output token price","unit":"usd","higher":false,"fileable":false,"measures":"Published output price"},{"slug":"gpqa-diamond","name":"GPQA Diamond","group":"Classic card","unit":"percent","higher":true,"measures":"Graduate science, multiple choice"},{"slug":"aime-2025","name":"AIME 2025","group":"Classic card","unit":"percent","higher":true,"measures":"Contest math, integer answers"},{"slug":"hle","name":"HLE, no tools","group":"Classic card","unit":"percent","higher":true,"measures":"Expert questions, text only"},{"slug":"swe-bench-verified","name":"SWE-bench Verified","group":"Classic card","unit":"percent","higher":true,"measures":"Real GitHub issues until the tests pass"},{"slug":"livecodebench","name":"LiveCodeBench","group":"Classic card","unit":"percent","higher":true,"measures":"Fresh contest programming, pass@1"}],"runs":[{"id":"6ac2aed41b362c4520f2121e","model":"Fable 5.1","maker":"Anthropic","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"fable-51","mark":"","test":"aa-briefcase-1-1","testName":"AA-Briefcase v1.1","score":1678,"labModel":"","labScore":null,"conditions":"Max","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f2121b","model":"Grok 4.7","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-47","mark":"","test":"aa-briefcase-1-1","testName":"AA-Briefcase v1.1","score":1657,"labModel":"","labScore":null,"conditions":"xHigh","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f2121c","model":"Grok 4.6","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-46","mark":"","test":"aa-briefcase-1-1","testName":"AA-Briefcase v1.1","score":1546,"labModel":"","labScore":null,"conditions":"High","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f2121d","model":"GPT-5.6 Sol","maker":"OpenAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"gpt-56","mark":"","test":"aa-briefcase-1-1","testName":"AA-Briefcase v1.1","score":1487,"labModel":"","labScore":null,"conditions":"Max","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2ab8394c3044477ff704c","model":"GPT-5.2 Thinking","maker":"OpenAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"","mark":"","test":"aime-2025","testName":"AIME 2025","score":100,"labModel":"","labScore":null,"conditions":"Lab published, no tools","tokPerSec":null,"notes":"https://openai.com/index/introducing-gpt-5-2/","published":true,"testedAt":"2025-12-11"},{"id":"6ac2ab8394c3044477ff7052","model":"Claude Opus 4.6","maker":"Anthropic","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"","mark":"","test":"aime-2025","testName":"AIME 2025","score":99.8,"labModel":"","labScore":null,"conditions":"Lab published, no tools, 5-trial average","tokPerSec":null,"notes":"Opus 4.6 system card prints 99.79%. Opus 4.7 moved on to USAMO.","published":true,"testedAt":"2026-02-05"},{"id":"6ac2ab8394c3044477ff704b","model":"GPT-5.6 Sol","maker":"OpenAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"","mark":"","test":"gpqa-diamond","testName":"GPQA Diamond","score":94.6,"labModel":"","labScore":null,"conditions":"Lab published","tokPerSec":null,"notes":"https://openai.com/index/gpt-5-6/","published":true,"testedAt":"2026-07-09"},{"id":"6ac2ab8394c3044477ff704f","model":"Claude Opus 4.7","maker":"Anthropic","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"","mark":"","test":"gpqa-diamond","testName":"GPQA Diamond","score":94.2,"labModel":"","labScore":null,"conditions":"Lab published, 10-trial average","tokPerSec":null,"notes":"Opus 4.7 system card, GPQA Diamond","published":true,"testedAt":"2026-04-16"},{"id":"6ac2f8479f1f269c91d419c8","model":"Qwen3.8 27B","maker":"Qwen","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"","mark":"","test":"livecodebench","testName":"LiveCodeBench","score":90.3,"labModel":"","labScore":null,"conditions":"LiveCodeBench v6, Qwen model card","tokPerSec":null,"notes":"https://huggingface.co/Qwen/Qwen3.8-27B","published":true,"testedAt":"2026-08-14"},{"id":"6ac2f8479f1f269c91d419c7","model":"Qwen3.8 27B","maker":"Qwen","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"","mark":"","test":"gpqa-diamond","testName":"GPQA Diamond","score":89.2,"labModel":"","labScore":null,"conditions":"Qwen model card","tokPerSec":null,"notes":"https://huggingface.co/Qwen/Qwen3.8-27B","published":true,"testedAt":"2026-08-14"},{"id":"6ac2ab8394c3044477ff7051","model":"Claude Opus 4.7","maker":"Anthropic","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"","mark":"","test":"swe-bench-verified","testName":"SWE-bench Verified","score":87.6,"labModel":"","labScore":null,"conditions":"Lab published","tokPerSec":null,"notes":"Opus 4.7 system card, SWE-bench Verified","published":true,"testedAt":"2026-04-16"},{"id":"6ac2ab8394c3044477ff704e","model":"GPT-5.2 Thinking","maker":"OpenAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"","mark":"","test":"swe-bench-verified","testName":"SWE-bench Verified","score":80,"labModel":"","labScore":null,"conditions":"Lab published","tokPerSec":null,"notes":"https://openai.com/index/introducing-gpt-5-2/","published":true,"testedAt":"2025-12-11"},{"id":"6ac2aed41b362c4520f21215","model":"GPT-5.6 Sol","maker":"OpenAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"gpt-56","mark":"","test":"deepswe-1-1","testName":"DeepSWE v1.1","score":72.7,"labModel":"","labScore":null,"conditions":"Max","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21213","model":"Grok 4.7","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-47","mark":"*","test":"deepswe-1-1","testName":"DeepSWE v1.1","score":71,"labModel":"","labScore":null,"conditions":"High effort","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21216","model":"Fable 5.1","maker":"Anthropic","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"fable-51","mark":"","test":"deepswe-1-1","testName":"DeepSWE v1.1","score":70,"labModel":"","labScore":null,"conditions":"Max","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21214","model":"Grok 4.6","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-46","mark":"","test":"deepswe-1-1","testName":"DeepSWE v1.1","score":65.2,"labModel":"","labScore":null,"conditions":"High","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21217","model":"Grok 4.7","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-47","mark":"","test":"eebench","testName":"EEBench","score":64,"labModel":"","labScore":null,"conditions":"xHigh","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f2122a","model":"Fable 5.1","maker":"Anthropic","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"fable-51","mark":"","test":"healthbench-pro","testName":"HealthBench Pro","score":62.1,"labModel":"","labScore":null,"conditions":"Max","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21229","model":"GPT-5.6 Sol","maker":"OpenAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"gpt-56","mark":"","test":"healthbench-pro","testName":"HealthBench Pro","score":60.5,"labModel":"","labScore":null,"conditions":"Max","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21222","model":"Fable 5.1","maker":"Anthropic","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"fable-51","mark":"","test":"terminal-bench-4","testName":"Terminal-Bench 4.0","score":57.9,"labModel":"","labScore":null,"conditions":"Max","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21227","model":"Grok 4.7","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-47","mark":"","test":"healthbench-pro","testName":"HealthBench Pro","score":56.7,"labModel":"","labScore":null,"conditions":"xHigh","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f2121a","model":"Fable 5.1","maker":"Anthropic","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"fable-51","mark":"","test":"eebench","testName":"EEBench","score":56.4,"labModel":"","labScore":null,"conditions":"Max","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21218","model":"Grok 4.6","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-46","mark":"","test":"eebench","testName":"EEBench","score":53,"labModel":"","labScore":null,"conditions":"High","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21212","model":"Fable 5.1","maker":"Anthropic","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"fable-51","mark":"","test":"cursorbench-4","testName":"CursorBench 4.0","score":51.8,"labModel":"","labScore":null,"conditions":"Max","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f2120e","model":"Fable 5.1","maker":"Anthropic","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"fable-51","mark":"","test":"price-out","testName":"$ per million","score":50,"labModel":"","labScore":null,"conditions":"Price on the Grok 4.7 card","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21228","model":"Grok 4.6","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-46","mark":"","test":"healthbench-pro","testName":"HealthBench Pro","score":48.5,"labModel":"","labScore":null,"conditions":"High","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2ab8394c3044477ff7050","model":"Claude Opus 4.7","maker":"Anthropic","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"","mark":"","test":"hle","testName":"HLE, no tools","score":46.9,"labModel":"","labScore":null,"conditions":"Lab published, no tools, max effort","tokPerSec":null,"notes":"Opus 4.7 system card, Humanity’s Last Exam","published":true,"testedAt":"2026-04-16"},{"id":"6ac2aed41b362c4520f2120f","model":"Grok 4.7","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-47","mark":"","test":"cursorbench-4","testName":"CursorBench 4.0","score":46.3,"labModel":"","labScore":null,"conditions":"xHigh","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2f8479f1f269c91d419c6","model":"Qwen3.8 27B","maker":"Qwen","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"qwen38-27b","mark":"†","test":"deepswe-1-1","testName":"DeepSWE v1.1","score":42.2,"labModel":"","labScore":null,"conditions":"Model card, Claude Code, temperature 1.0, 256K","tokPerSec":null,"notes":"https://huggingface.co/Qwen/Qwen3.8-27B","published":true,"testedAt":"2026-08-14"},{"id":"6ac2aed41b362c4520f21211","model":"GPT-5.6 Sol","maker":"OpenAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"gpt-56","mark":"","test":"cursorbench-4","testName":"CursorBench 4.0","score":41.7,"labModel":"","labScore":null,"conditions":"Max","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21210","model":"Grok 4.6","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-46","mark":"","test":"cursorbench-4","testName":"CursorBench 4.0","score":40.4,"labModel":"","labScore":null,"conditions":"High","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21219","model":"GPT-5.6 Sol","maker":"OpenAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"gpt-56","mark":"","test":"eebench","testName":"EEBench","score":39.4,"labModel":"","labScore":null,"conditions":"Max","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f2121f","model":"Grok 4.7","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-47","mark":"","test":"terminal-bench-4","testName":"Terminal-Bench 4.0","score":37.6,"labModel":"","labScore":null,"conditions":"xHigh","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21221","model":"GPT-5.6 Sol","maker":"OpenAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"gpt-56","mark":"","test":"terminal-bench-4","testName":"Terminal-Bench 4.0","score":37.3,"labModel":"","labScore":null,"conditions":"Max","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2ab8394c3044477ff704d","model":"GPT-5.2 Thinking","maker":"OpenAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"","mark":"","test":"hle","testName":"HLE, no tools","score":34.5,"labModel":"","labScore":null,"conditions":"Lab published, no tools","tokPerSec":null,"notes":"https://openai.com/index/introducing-gpt-5-2/","published":true,"testedAt":"2025-12-11"},{"id":"6ac2aed41b362c4520f21220","model":"Grok 4.6","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-46","mark":"","test":"terminal-bench-4","testName":"Terminal-Bench 4.0","score":20.3,"labModel":"","labScore":null,"conditions":"High","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f2120d","model":"GPT-5.6 Sol","maker":"OpenAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"gpt-56","mark":"","test":"price-out","testName":"$ per million","score":20,"labModel":"","labScore":null,"conditions":"Price on the Grok 4.7 card","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21223","model":"Grok 4.7","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-47","mark":"","test":"harvey-legal","testName":"Harvey Legal Agent","score":19.6,"labModel":"","labScore":null,"conditions":"xHigh","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21224","model":"Grok 4.6","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-46","mark":"","test":"harvey-legal","testName":"Harvey Legal Agent","score":15.8,"labModel":"","labScore":null,"conditions":"High","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f2120a","model":"Fable 5.1","maker":"Anthropic","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"fable-51","mark":"","test":"price-in","testName":"$ per million","score":10,"labModel":"","labScore":null,"conditions":"Price on the Grok 4.7 card","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21226","model":"Fable 5.1","maker":"Anthropic","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"fable-51","mark":"","test":"harvey-legal","testName":"Harvey Legal Agent","score":6.7,"labModel":"","labScore":null,"conditions":"Max","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f2120b","model":"Grok 4.7","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-47","mark":"","test":"price-out","testName":"$ per million","score":6,"labModel":"","labScore":null,"conditions":"Published API price","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f2120c","model":"Grok 4.6","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-46","mark":"","test":"price-out","testName":"$ per million","score":6,"labModel":"","labScore":null,"conditions":"Published API price","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21209","model":"GPT-5.6 Sol","maker":"OpenAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"gpt-56","mark":"","test":"price-in","testName":"$ per million","score":4,"labModel":"","labScore":null,"conditions":"Price on the Grok 4.7 card","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21225","model":"GPT-5.6 Sol","maker":"OpenAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"gpt-56","mark":"","test":"harvey-legal","testName":"Harvey Legal Agent","score":2.5,"labModel":"","labScore":null,"conditions":"Max","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21207","model":"Grok 4.7","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-47","mark":"","test":"price-in","testName":"$ per million","score":2,"labModel":"","labScore":null,"conditions":"Published API price","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"},{"id":"6ac2aed41b362c4520f21208","model":"Grok 4.6","maker":"xAI","params":"","quant":"","cardClass":"lab","cardLabel":"Lab published","kind":"published","column":"grok-46","mark":"","test":"price-in","testName":"$ per million","score":2,"labModel":"","labScore":null,"conditions":"Published API price","tokPerSec":null,"notes":"https://x.ai/news/grok-4-7","published":true,"testedAt":"2026-09-21"}]}