GPT-6.2 Astra (max): benchmark scores, price and ranking

GPT-6.2 Astra (max) from OpenAI ranks #7 of 827 LLMs on the LLMs Tiger aggregate benchmark score (60.5), based on 11 benchmark results captured up to 2026-11-18.

Organisation
OpenAI
Type
Frontier model
Open weights
No
Released
2026-11-17 (source: epoch)
Cheapest API price
$8.00 in / $40.00 out per million tokens (via azure, observed 2026-11-18T15:47Z)
Aggregate rank
#7 of 827 (score 60.5)

Relative to other models, GPT-6.2 Astra (max) is strongest on GPQA Diamond (#1 of 302) and weakest on MirrorCode (#8 of 35).

Across the 11 benchmarks where at least ten models are scored, GPT-6.2 Astra (max) is in the top 10% on 10 and in the top half on 11.

Its cheapest listed input price, $8.00 per million tokens, is lower than that of 6% of the 443 models with a listed price.

GPT-6.2 Astra (max) benchmark results

GPT-6.2 Astra (max) benchmark scores, ranked against all models measured on each benchmark
BenchmarkScoreRankSourceCaptured
GPQA Diamond96.8%#1 of 302Epoch AI Benchmarking Hub2026-11-18
ARC-AGI-295.8%#1 of 97ARC Prize official leaderboard2026-11-18
Mystery Games95.0%#1 of 69Epoch AI Benchmarking Hub2026-11-18
SimpleQA Verified78.5%#2 of 83Epoch AI Benchmarking Hub2026-11-18
Chess Puzzles78.0%#3 of 84Epoch AI Benchmarking Hub2026-11-18
EBR-Bench92.4%#1 of 27Epoch AI Benchmarking Hub2026-11-18
FrontierMath T4100.0%#3 of 60Epoch AI Benchmarking Hub2026-11-18
FrontierMath T1-392.6%#4 of 78Epoch AI Benchmarking Hub2026-11-18
OTIS Mock AIME100.0%#11 of 206Epoch AI Benchmarking Hub2026-11-18
Furniture Assembly89.2%#2 of 36Epoch AI Benchmarking Hub2026-11-18
MirrorCode67.8%#8 of 35Epoch AI Benchmarking Hub2026-11-18

About GPT-6.2 Astra (max)

GPT-6.2 Astra (max) is a frontier model from OpenAI. It was released on 2026-11-17, according to Epoch AI. The cheapest API listing we track charges $8.00 per million input tokens and $40.00 per million output tokens, through azure. We hold 11 benchmark results for it across 2 areas, and 10 are in the top tenth of their boards.

Where GPT-6.2 Astra (max) stands

On our aggregate score, GPT-6.2 Astra (max) is #7 of 827 ranked models, in the top ten. Its score is 60.5; the scale is centred on 50, so it sits 10.5 points above the typical model. That beats 99% of the ranked field. Among the 91 ranked OpenAI models it is #2. The next one up is GPT-6 Astra (max) at 62.3. The next one down is GPT-6.1 Sol (max) at 60.4.

GPT-6.2 Astra (max), benchmark by benchmark

GPQA Diamond is where GPT-6.2 Astra (max) does best: 96.8%, first of 302. The next result is Gemini 4 Carbon (high) at 96.5%, 0.3 points behind. It covers expert-level science questions. The board median is 75.1%, so GPT-6.2 Astra (max) is 21.7 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-11-18.

ARC-AGI-2 is where GPT-6.2 Astra (max) does best: 95.8%, first of 97. The next result is GPT-6 Astra (max) at 95.0%, 0.8 points behind. It covers abstract reasoning, ARC Prize public eval set. The board median is 40.1%, so GPT-6.2 Astra (max) is 55.7 points above it. Source: ARC Prize official leaderboard, captured 2026-11-18.

Mystery Games is where GPT-6.2 Astra (max) does best: 95.0%, first of 69. The next result is Claude Fable 5.2 (max) at 90.0%, 5.0 points behind. It covers deductive reasoning puzzles. The board median is 26.0%, so GPT-6.2 Astra (max) is 69.0 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-11-18.

On SimpleQA Verified, GPT-6.2 Astra (max) scores 78.5% and ranks #2 of 83. Only one model scores higher, led by Claude Fable 5.2 (max) at 94.0%, which is 15.5 points ahead. It covers short factual questions. The board median is 41.4%, so GPT-6.2 Astra (max) is 37.1 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-11-18.

On Chess Puzzles, GPT-6.2 Astra (max) scores 78.0% and ranks #3 of 84. Only 2 models score higher, led by Claude Fable 5.2 (max) at 79.0%, which is 1.0 points ahead. It covers chess tactics puzzles. The board median is 33.0%, so GPT-6.2 Astra (max) is 45.0 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-11-18.

EBR-Bench is where GPT-6.2 Astra (max) does best: 92.4%, first of 27. The next result is Claude Fable 5.2 (max) at 90.5%, 1.9 points behind. It covers evidence-based reasoning. The board median is 39.5%, so GPT-6.2 Astra (max) is 52.9 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-11-18.

On FrontierMath T1-3, GPT-6.2 Astra (max) scores 92.6% and ranks #4 of 78. Only 3 models score higher, led by Claude Fable 5.2 (max) at 95.1%, which is 2.5 points ahead. It covers research-level maths, tiers 1-3. The board median is 61.6%, so GPT-6.2 Astra (max) is 31.1 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-11-18.

On OTIS Mock AIME, GPT-6.2 Astra (max) scores 100.0% and ranks #11 of 206. Only 10 models score higher, led by Claude Fable 5.1 (max) at 100.0%, which is 0.0 points ahead. It covers competition maths (AIME 2024-25 style). The board median is 66.4%, so GPT-6.2 Astra (max) is 33.6 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-11-18.

On Furniture Assembly, GPT-6.2 Astra (max) scores 89.2% and ranks #2 of 36. Only one model scores higher, led by Claude Fable 5.2 (max) at 90.0%, which is 0.8 points ahead. It covers planning a physical assembly. The board median is 44.2%, so GPT-6.2 Astra (max) is 45.0 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-11-18.

MirrorCode puts GPT-6.2 Astra (max) at 67.8%, #8 of 35. That beats 77% of the models tested, but 30.0 points short of Claude Sonnet 5.5 (max) (97.8%). The results either side are Claude Fable 5 (max) (70.0%) and deepseek-v4.1-flash-max (64.4%). It covers reimplementing programs from behaviour. The board median is 48.9%, so GPT-6.2 Astra (max) is 18.9 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-11-18.

The full table above also lists the other 1 result, including boards with fewer than ten models, where a rank says little.

Strengths and weaknesses

Reasoning, maths and science: strong, ahead of 97% of the field on average across 5 benchmarks. Its best board here is GPQA Diamond (#1 of 302) and its weakest is OTIS Mock AIME (#11 of 206). General knowledge and chat: strong, ahead of 93% of the field on average across 6 benchmarks. Its best board here is Mystery Games (#1 of 69) and its weakest is MirrorCode (#8 of 35).

Price and value

The median input price across the 443 priced models is $0.482 per million tokens, so GPT-6.2 Astra (max) costs more than the typical listing. Models that score within 1.5 points or better and cost less: GPT-6.1 Terra (max) ($1.00, 59.8), Gemini 4 Carbon (high) ($2.00, 62.4), Claude Sonnet 5.5 (max) ($2.00, 60.7). Similar scores cost more on Claude Fable 5.1 (max) ($10.00).

Release history

The OpenAI model released before it was GPT-6.1 Terra (max) on 2026-10-26, which scores 59.8. The next one was GPT-6.2 Sol (Max) on 2026-11-17, at 52.3.

Reading these numbers

The 11 scores on this page come from 2 sources: Epoch AI Benchmarking Hub (10), ARC Prize official leaderboard (1). 1 of them come from an official leaderboard and 10 from datasets or third-party runs. The leaderboard also lists 5 other reasoning-effort settings of this model, which redirect here. This page shows the highest.

GPT-6.2 Astra (max) is also listed at 5 other reasoning-effort settings; this page shows the highest-effort row, which is the one the leaderboard keeps by default.

Models ranked nearby

More from OpenAI

Compare it on the live leaderboard → · How the aggregate score works