{"id":"gpt-6-2-astra","slug":"gpt-6-2-astra","name":"GPT-6.2 Astra","orgId":"openai","family":"gpt-6-2-astra","releaseDate":"2026-11-17","availability":"ga","openWeights":false,"license":"Proprietary","paramsTotal":null,"paramsSource":null,"contextWindow":1050000,"knowledgeCutoff":null,"modelCardUrl":null,"hfRepo":null,"modalities":["text","file","image"],"aliases":["gpt-6.2-astra_max","GPT-6.2 Astra","openai/gpt-6.2-astra","OpenAI: GPT-6.2 Astra","gpt-6.2-astra","gpt 6.2 astra","gpt-6-2-astra","GPT-6.2 Astra (Max)","gpt-6.2-astra_high","gpt-6.2-astra_xhigh","gpt-6.2-astra_low","gpt-6.2-astra_medium"],"aliasNote":null,"contextSource":"openrouter","tracked":{"trackedSince":"2026-11-17","announced":"2026-11-17","blurb":"OpenAI's GPT-6.2 Astra model for hard professional work, coding, and complex projects. It takes text and image input, has a 1,050,000-token context window, and is live in the API as gpt-6.2-astra. Standard short-context prices are $8 input, $0.40 cached input, and $40 output per 1M tokens — 20% lower input and output token prices than GPT-6 Astra, with a 60% discount on cached input. Also available in ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu. Fast and Ultrafast are processing modes of this model, not separate models; official API docs price Ultrafast at 3 times the Standard rate through February 28, 2027, and 6 times the Standard rate starting March 1, 2027 (Pacific time). Supports computer use and other tools through the Responses API. No open weights.","api":"live","apiModelId":"gpt-6.2-astra","openWeights":"no","facts":[{"label":"Input / output","value":"Text and image in, text out"},{"label":"Context","value":"1,050,000 tokens (up to 922,000 input), up to 128,000 output tokens"},{"label":"Knowledge cutoff","value":"May 18, 2026"},{"label":"Price (OpenAI API)","value":"Standard short-context: $8 input, $0.40 cached input, $40 output per 1M tokens (20% lower input and output than GPT-6 Astra, with a 60% discount on cached input). Cache writes $10 per 1M tokens for 30-minute retention ($20 for 12-hour). Prompts over 272K input tokens: input and cache rates double and output is 1.5 times the short-context rate for the full request. Batch is 50% of Standard. Fast mode is twice the Standard rate. Ultrafast is 3 times the Standard rate through February 28, 2027, and 6 times the Standard rate starting March 1, 2027 (Pacific time)"},{"label":"Reasoning effort","value":"low, medium, high, xhigh and max (the API rejects none)"},{"label":"Endpoints","value":"Responses, Chat Completions and Batch; tool calling requires the Responses API. Official tools include computer use, web search, file search, code interpreter, and others"},{"label":"Available","value":"OpenAI API as gpt-6.2-astra; ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu. Pro 500 subscribers can use Ultrafast with this model; eligible Enterprise and Edu workspaces after an admin enables it"},{"label":"Modes","value":"Fast and Ultrafast are processing tiers of this API model, not separate model pages"},{"label":"Preparedness (OpenAI)","value":"Official system card: Critical threshold for cybersecurity and High threshold for biological and chemical capabilities; first externally released OpenAI model to meet the High threshold for AI Self-improvement"},{"label":"Weights","value":"Not released"},{"label":"Size","value":"Not stated"},{"label":"Related","value":"Separate from GPT-6 Astra, GPT-6 Astra Law, GPT-6.2 Sol, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Luna, GPT-6 Cyber, and GPT-6 Rosalind. The API pricing page also lists gpt-6.2-astra-law (GPT-6.2 Astra for Law), which is not this model and is not tracked here"}],"claims":["Lab claim: self-reported official blog scores on Agents' Last Exam of 5.2 points higher than GPT-6 Astra at less than a third of the cost, and 5.5 points ahead of Claude Fable 5.2 at 5% of the cost. The official blog notes that leaderboard results for Fable 5.2 exclude nine tasks that used a fallback model.","Lab claim: self-reported official blog scores on Terminal-Bench-Science 0.1 of 9 percentage points higher than GPT-6 Astra at less than one-third the cost."],"pitch":"a model for your hardest projects","peer":{"slug":"gpt-6-2-sol","label":"GPT-6.2 Sol"},"sources":[{"kind":"lab","title":"Introducing GPT-6.2 Astra and Sol (OpenAI blog)","url":"https://openai.com/index/introducing-gpt-6-2-astra-sol/","date":"2026-11-17"},{"kind":"lab","title":"GPT-6.2 Astra model page (OpenAI API docs)","url":"https://developers.openai.com/api/docs/models/gpt-6.2-astra","date":"2026-11-17"},{"kind":"lab","title":"OpenAI API pricing","url":"https://developers.openai.com/api/docs/pricing","date":"2026-11-17"},{"kind":"lab","title":"OpenAI API changelog","url":"https://developers.openai.com/api/docs/changelog","date":"2026-11-17"},{"kind":"lab","title":"GPT-6.2 Astra and Sol system card","url":"https://deploymentsafety.openai.com/gpt-6-2-astra-sol/gpt-6-2-astra-sol.pdf","date":"2026-11-17"},{"kind":"lab","title":"OpenAI on X: GPT-6.2 Astra and Sol","url":"https://x.com/OpenAI/status/2122750124543926515","date":"2026-11-17"},{"kind":"lab","title":"OpenAIDevs on X: GPT-6.2 Astra and Sol","url":"https://x.com/OpenAIDevs/status/2122752500877144461","date":"2026-11-17"}],"firstBenchmark":{"date":"2026-11-17","benchmarkId":"gpqa-diamond"}},"listedSince":"2026-11-18","organization":{"id":"openai","name":"OpenAI","slug":"openai","country":null,"url":null,"color":"#10A37F","mark":"O"},"price":{"modelId":"gpt-6-2-astra","provider":"openai","inputPerM":null,"outputPerM":null,"blendedPerM":16.000000000000004,"tokensPerS":null,"observedAt":"2026-11-22","sourceId":"openrouter"},"answer":{"word":"Too new.","scope":"Too new to call. The crown still needs confirming evidence.","lead":"As of Nov 22, 2026, GPT-6.2 Astra is leading, crown pending confirmation in Science (since Nov 22, 2026; sources MysteryMechanism; methodology v1.0). Too new to call Science. Ranked #1. A crown change still needs two confirming signals. That board does not award a crown.","confidence":"low"},"status":"BENCHMARKED","verdicts":[{"id":"v_overall_general-llm_gpt-6-2-astra_2026-11-22_open_RANKED","model":"gpt-6-2-astra","model_name":"GPT-6.2 Astra","category":"general-llm","category_name":"General LLM","slice":"overall","state":"RANKED","rank":46,"present_evidence":[],"missing_evidence":[],"coverage":null,"segments":[],"confirm":[],"margin":27.485380116959064,"confidence":"medium","valid_from":"2026-11-22","valid_to":null,"configuration":null,"rationale":"Ranked #46 in General LLM.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#46 in General LLM"},{"id":"v_pareto_general-llm_gpt-6-2-astra_2026-11-22_open_RANKED","model":"gpt-6-2-astra","model_name":"GPT-6.2 Astra","category":"general-llm","category_name":"General LLM","slice":"pareto","state":"RANKED","rank":46,"present_evidence":[],"missing_evidence":[],"coverage":null,"segments":[],"confirm":[],"margin":27.485380116959064,"confidence":"medium","valid_from":"2026-11-22","valid_to":null,"configuration":null,"rationale":"Ranked #46 in General LLM on best per dollar.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#46 in General LLM (best per dollar)"},{"id":"v_overall_reasoning_gpt-6-2-astra_2026-11-22_open_RANKED","model":"gpt-6-2-astra","model_name":"GPT-6.2 Astra","category":"reasoning","category_name":"Reasoning","slice":"overall","state":"RANKED","rank":54,"present_evidence":[],"missing_evidence":[],"coverage":null,"segments":[],"confirm":[],"margin":32.352941176470594,"confidence":"medium","valid_from":"2026-11-22","valid_to":null,"configuration":null,"rationale":"Ranked #54 in Reasoning.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#54 in Reasoning"},{"id":"v_pareto_reasoning_gpt-6-2-astra_2026-11-22_open_RANKED","model":"gpt-6-2-astra","model_name":"GPT-6.2 Astra","category":"reasoning","category_name":"Reasoning","slice":"pareto","state":"RANKED","rank":54,"present_evidence":[],"missing_evidence":[],"coverage":null,"segments":[],"confirm":[],"margin":32.352941176470594,"confidence":"medium","valid_from":"2026-11-22","valid_to":null,"configuration":null,"rationale":"Ranked #54 in Reasoning on best per dollar.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#54 in Reasoning (best per dollar)"},{"id":"v_overall_science_gpt-6-2-astra_2026-11-22_open_PENDING","model":"gpt-6-2-astra","model_name":"GPT-6.2 Astra","category":"science","category_name":"Science","slice":"overall","state":"PENDING","rank":1,"present_evidence":[],"missing_evidence":[],"coverage":null,"segments":[],"confirm":[],"margin":null,"confidence":"low","valid_from":"2026-11-22","valid_to":null,"configuration":"max","rationale":"Too new to call Science. Ranked #1. A crown change still needs two confirming signals. That board does not award a crown.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"Pending · Science"},{"id":"v_overall_coding_gpt-6-2-astra_2026-11-22_open_INSUFFICIENT_DATA","model":"gpt-6-2-astra","model_name":"GPT-6.2 Astra","category":"coding","category_name":"Coding","slice":"overall","state":"INSUFFICIENT_DATA","rank":null,"present_evidence":[{"benchmark":"lmarena-text-coding","name":"LMArena Text (style-controlled, coding)","rank":32}],"missing_evidence":[{"benchmark":"terminal-bench-4","name":"Terminal-Bench 4.0"},{"benchmark":"swe-bench-pro-hard","name":"SWE-bench Pro V2 HARD"}],"coverage":null,"segments":[],"confirm":[],"margin":null,"confidence":"low","valid_from":"2026-11-22","valid_to":null,"configuration":null,"rationale":"#32 on LMArena Text (style-controlled, coding); missing Terminal-Bench 4.0 and SWE-bench Pro V2 HARD","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"Not enough data"},{"id":"v_overall_coding-web-dev_gpt-6-2-astra_2026-11-22_open_PENDING","model":"gpt-6-2-astra","model_name":"GPT-6.2 Astra","category":"coding-web-dev","category_name":"Coding › Web dev","slice":"overall","state":"PENDING","rank":4,"present_evidence":[],"missing_evidence":[],"coverage":null,"segments":[],"confirm":[],"margin":2.919708029197082,"confidence":"low","valid_from":"2026-11-22","valid_to":null,"configuration":null,"rationale":"Too new to call Coding › Web dev. A crown change still needs two confirming signals.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"Pending · Coding › Web dev"},{"id":"v_pareto_coding-web-dev_gpt-6-2-astra_2026-11-22_open_NEAR_SOTA","model":"gpt-6-2-astra","model_name":"GPT-6.2 Astra","category":"coding-web-dev","category_name":"Coding › Web dev","slice":"pareto","state":"NEAR_SOTA","rank":4,"present_evidence":[],"missing_evidence":[],"coverage":null,"segments":[],"confirm":[],"margin":2.919708029197082,"confidence":"medium","valid_from":"2026-11-22","valid_to":null,"configuration":null,"rationale":"Close on Coding › Web dev (best per dollar): rank #4.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"Near-SOTA · Coding › Web dev (best per dollar)"},{"id":"v_overall_long-context_gpt-6-2-astra_2026-11-22_open_RANKED","model":"gpt-6-2-astra","model_name":"GPT-6.2 Astra","category":"long-context","category_name":"Long context","slice":"overall","state":"RANKED","rank":49,"present_evidence":[],"missing_evidence":[],"coverage":null,"segments":[],"confirm":[],"margin":29.411764705882348,"confidence":"medium","valid_from":"2026-11-22","valid_to":null,"configuration":null,"rationale":"Ranked #49 in Long context.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#49 in Long context"},{"id":"v_pareto_long-context_gpt-6-2-astra_2026-11-22_open_RANKED","model":"gpt-6-2-astra","model_name":"GPT-6.2 Astra","category":"long-context","category_name":"Long context","slice":"pareto","state":"RANKED","rank":49,"present_evidence":[],"missing_evidence":[],"coverage":null,"segments":[],"confirm":[],"margin":29.411764705882348,"confidence":"medium","valid_from":"2026-11-22","valid_to":null,"configuration":null,"rationale":"Ranked #49 in Long context on best per dollar.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#49 in Long context (best per dollar)"},{"id":"v_overall_vision_gpt-6-2-astra_2026-11-22_open_INSUFFICIENT_DATA","model":"gpt-6-2-astra","model_name":"GPT-6.2 Astra","category":"vision","category_name":"Vision","slice":"overall","state":"INSUFFICIENT_DATA","rank":null,"present_evidence":[{"benchmark":"gdp-pdf","name":"GDP.pdf","rank":12}],"missing_evidence":[{"benchmark":"lmarena-vision","name":"LMArena Vision (style-controlled)"}],"coverage":null,"segments":[],"confirm":[],"margin":null,"confidence":"low","valid_from":"2026-11-22","valid_to":null,"configuration":"max","rationale":"#12 on GDP.pdf; missing LMArena Vision (style-controlled)","safeguard_flags":["off_board"],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"Not enough data"},{"id":"v_overall_factuality_gpt-6-2-astra_2026-11-22_open_RANKED","model":"gpt-6-2-astra","model_name":"GPT-6.2 Astra","category":"factuality","category_name":"Factuality","slice":"overall","state":"RANKED","rank":89,"present_evidence":[],"missing_evidence":[],"coverage":null,"segments":[],"confirm":[],"margin":52.352941176470594,"confidence":"medium","valid_from":"2026-11-22","valid_to":null,"configuration":null,"rationale":"Ranked #89 in Factuality.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#89 in Factuality"},{"id":"v_pareto_factuality_gpt-6-2-astra_2026-11-22_open_RANKED","model":"gpt-6-2-astra","model_name":"GPT-6.2 Astra","category":"factuality","category_name":"Factuality","slice":"pareto","state":"RANKED","rank":89,"present_evidence":[],"missing_evidence":[],"coverage":null,"segments":[],"confirm":[],"margin":52.352941176470594,"confidence":"medium","valid_from":"2026-11-22","valid_to":null,"configuration":null,"rationale":"Ranked #89 in Factuality on best per dollar.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#89 in Factuality (best per dollar)"},{"id":"v_overall_arc_gpt-6-2-astra_2026-11-22_open_PENDING","model":"gpt-6-2-astra","model_name":"GPT-6.2 Astra","category":"arc","category_name":"Abstraction (ARC)","slice":"overall","state":"PENDING","rank":1,"present_evidence":[],"missing_evidence":[],"coverage":null,"segments":[],"confirm":[],"margin":null,"confidence":"low","valid_from":"2026-11-22","valid_to":null,"configuration":"medium","rationale":"Too new to call Abstraction (ARC). Ranked #1. A crown change still needs two confirming signals. That board does not award a crown.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"Pending · Abstraction (ARC)"},{"id":"v_overall_legal_gpt-6-2-astra_2026-11-22_open_RANKED","model":"gpt-6-2-astra","model_name":"GPT-6.2 Astra","category":"legal","category_name":"Legal","slice":"overall","state":"RANKED","rank":16,"present_evidence":[],"missing_evidence":[],"coverage":null,"segments":[],"confirm":[],"margin":null,"confidence":"low","valid_from":"2026-11-22","valid_to":null,"configuration":"max","rationale":"Ranked #16 in Legal. That board does not award a crown.","safeguard_flags":[],"co_sota_with":[],"challengers":[],"methodology":"1.0","label":"#16 in Legal"}]}