Is GPT-6.2 Astra state of the art?Too new.Too new to call. The crown still needs confirming evidence.
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.
Facts and scores
GPT-6.2 Astra is a closed-weights text, file and image model from OpenAI, listed on GPQA Diamond since Nov 17, 2026; no public parameter count, a 1,050K-token context window and $16.00 per 1M tokens blended.
How it scores
- On GPQA Diamond it scores 96.8% (CI 94.4%–99.2%), #1 of 245 (top 1%); level with Gemini 4 Hydrogen below.
- On SimpleQA Verified it scores 78.5% (CI 76.0%–81.0%), #2 of 99 (top 3%); 15.5 pts behind Claude Fable 5.2; Claude Fable 5.2 is 15.5 pts above and Gemini 4 Carbon is 0.8 pts below.
- On LMArena WebDev it scores 1766 (CI 1751–1781), #5 of 165 (top 4%); 58 behind Claude Fable 5.2; Gemini 4 Carbon is 3 above and Gemini 4 Neon is 1 below.
History
First result on isitsota: Nov 17, 2026. Best overall rank so far: #1 in Science, first held Nov 22, 2026.
The ruling, in plain English
Ranked #46 in General LLM.
Evidence
| Source | Score | Date | Type | Receipt |
|---|---|---|---|---|
| LMArena Text (style-controlled, overall) GPT-6.2 Astra (Max) | 1468 1456–1481 | 2026-11-19 | INDEP. | snap_280a9b5ca493 |
| LMArena Text (style-controlled, hard prompts) GPT-6.2 Astra (Max) | 1485 1469–1501 | 2026-11-19 | INDEP. | snap_280a9b5ca493 |
| MysteryMechanism openai/gpt-6.2-astra · max · Source: Vals AI | 53.2 46.6–59.7 | 2026-11-21 | INDEP. | snap_vals_54c41eb6660d |
| LMArena Text (style-controlled, coding) GPT-6.2 Astra (Max) | 1523 1495–1550 | 2026-11-19 | INDEP. | snap_280a9b5ca493 |
| LMArena WebDev GPT-6.2 Astra (Max) | 1766 1751–1781 | 2026-11-20 | INDEP. | snap_2f4a3ebc7e46 |
| LMArena Text (style-controlled, longer query) GPT-6.2 Astra (Max) | 1476 1456–1495 | 2026-11-19 | INDEP. | snap_280a9b5ca493 |
| GDP.pdf gpt-6.2-astra_max · max | 49.2% | 2026-11-17 | INDEP. | snap_2d63325b3124 |
| LMArena Text factuality GPT-6.2 Astra (Max) | 1439 1428–1450 | 2026-11-19 | INDEP. | snap_275d95909246 |
| ARC-AGI gpt-6.2-astra_medium · medium | 98.5% | 2026-11-17 | INDEP. | snap_2d63325b3124 |
| Legal Research Bench openai/gpt-6.2-astra · max · Source: Vals AI | 42.8 36.0–49.5 | 2026-11-21 | INDEP. | snap_vals_e4530282a8b5 |
Rows marked Vals AI come from Vals AI's proprietary benchmarks, republished with permission. Each receipt links to the benchmark on Vals.
Where it sits in the succession
Crown intervals for this model are still empty.
Where it is not the crown
| Category | This model | Rank | Leader |
|---|---|---|---|
| General LLM | #46 in this category | #46 | Gemini 4 Carbon |
| Reasoning | #54 in this category | #54 | Gemini 4 Carbon |
| Coding | Not enough data | — | Claude Sonnet 5.5 |
| Long context | #49 in this category | #49 | Gemini 4 Argon |
| Vision | Not enough data | — | — |
| Factuality | #89 in this category | #89 | Gemini 4 Carbon |
| Legal | #16 in this category | #16 | Gemini 4 Carbon |
Related pages
- General LLM
- Reasoning
- Science
- Coding
- Coding › Web dev
- Long context
- Vision
- Factuality
- Abstraction (ARC)
- Legal
- Compare with the current leader in General LLM: Gemini 4 Carbon
- Compare with the current leader in General LLM: Gemini 4 Argon
- Compare with the current leader in Reasoning: Gemini 4 Carbon
- Compare with the current leader in Reasoning: Gemini 4 Argon
- Compare with the current leader in Reasoning: Gemini 4 Neon
- Compare with the current leader in Coding › Web dev: Claude Fable 5.2
- Compare with the current leader in Coding › Web dev: Claude Opus 5.5
- Compare with the current leader in Long context: Gemini 4 Argon
- Compare with the current leader in Long context: Gemini 4 Carbon
- Compare with the current leader in Long context: Gemini 4 Neon
- Compare with the current leader in Factuality: Gemini 4 Carbon
- Compare with the current leader in Factuality: Gemini 4 Neon
- GPT-6.2 Astra vs Claude Fable 5
- GPT-6.2 Astra vs Claude Opus 5
- GPT-6.2 Astra vs Claude Sonnet 5.5
- MysteryMechanism
- Legal Research Bench
- LMArena WebDev
- LMArena Text (style-controlled, overall)
- LMArena Text (style-controlled, coding)
- LMArena Text (style-controlled, hard prompts)
- LMArena Text (style-controlled, longer query)
- LMArena Text factuality
- ARC-AGI
- GDP.pdf
- Lab: OpenAI
- Family: GPT-6
Is GPT-6.2 Astra SOTA?
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).
What is GPT-6.2 Astra best at?
As of Nov 22, 2026, GPT-6.2 Astra is leading, crown pending confirmation in Science (since Nov 22, 2026; sources MysteryMechanism).
How does GPT-6.2 Astra rank in Coding?
As of Nov 22, 2026, GPT-6.2 Astra is short of the evidence for a rank in Coding (since Nov 22, 2026; sources LMArena Text (style-controlled, coding); methodology v1.0).