Benchmark Almanac

METR Capability Index

GPT-6 Astra 859.0 leads since November 2026
Not in model cards

METR fits an item response model to agent results on Time Horizon 1.1, TH2 and MirrorCode. The index places models on a common ability scale.

METR publishes model estimates and uncertainty intervals alongside its November 2026 capability report.

Made by
METR
Released
November 2026
Kind
Task length
Level
Research frontier
Scored
Checked automatically
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Code → Text, Actions
Interaction
Agentic
Environment
Terminal
Items from
Real work and use
Run by
Anyone
In views
Agents
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
September 2026
Scores refreshed
November 21st, 2026
30
Entries on its own leaderboard, 28 models we track

GPT-6 Astra leads with 859.0 among models released by September 2026 GPT-6 Astra leads with 859.0 among models released by September 2026

Best score per entry from the benchmark's own leaderboard, read on November 21st, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

The newest model with a score here came out in September 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 GPT-6 Astra 859.0
  2. 2 Claude Opus 5.5 854.1
  3. 3 Claude Fable 5.1 839.1
  4. 4 Claude Opus 5 796.2
  5. 5 Claude Fable 5 778.7
  6. 6 GPT-5.6 Sol 738.3
  7. 7 Claude Opus 4.8 708.6
  8. 8 GPT-5.5 705.7
  9. 9 Claude Mythos Preview (early) 671.2
  10. 10 Claude Opus 4.6 633.0
  11. 11 Claude Opus 4.7 631.2
  12. 12 GPT-5.4 593.2
  13. 13 Gemini 3.1 Pro 568.9
  14. 14 GPT-5.3 Codex 566.5
  15. 15 GPT-5.2 551.5
  16. 16 Claude Opus 4.5 533.3
  17. 17 Gemini 3 Pro 490.5
  18. 18 GPT-5 478.6
  19. 19 GPT-5.1-Codex-Max 477.7
  20. 20 o3 417.3
  21. 21 Claude Opus 4.1 398.5
  22. 22 Claude Opus 4 392.6
  23. 23 Claude 3.7 Sonnet 310.0
  24. 24 o1 237.4
  25. 25 Claude 3.5 Sonnet (October 2024) 192.1
  26. 26 o1-preview 140.2
  27. 27 Claude 3.5 Sonnet (Old) 70.6
  28. 28 GPT-4o May 2024 0.0
  29. 29 Claude 3 Opus -84.5
  30. 30 GPT-4 Turbo Apr 2024 -105.6

Source: Official METR Capability Index leaderboard

#EntryScore ±SettingReleasedRun onSource
1 GPT-6 Astra 859.0 761.1 to 1085.3 Sep 2026 Sep 2026 Official leaderboard
2 Claude Opus 5.5 854.1 765.2 to 1069.3 Sep 2026 Sep 2026 Official leaderboard
3 Claude Fable 5.1 839.1 753.7 to 1019.1 Sep 2026 Sep 2026 Official leaderboard
4 Claude Opus 5 796.2 735.9 to 910.1 Jul 2026 Jul 2026 Official leaderboard
5 Claude Fable 5 778.7 712.9 to 890.9 Jun 2026 Jun 2026 Official leaderboard
6 GPT-5.6 Sol 738.3 681.6 to 855.8 Jul 2026 Jul 2026 Official leaderboard
7 Claude Opus 4.8 708.6 661.3 to 784.3 May 2026 May 2026 Official leaderboard
8 GPT-5.5 705.7 657.8 to 817.1 Apr 2026 Apr 2026 Official leaderboard
9 Claude Mythos Preview (early) 671.2 637.5 to 761.4 Apr 2026 Official leaderboard
10 Claude Opus 4.6 633.0 Feb 2026 Feb 2026 Official leaderboard
11 Claude Opus 4.7 631.2 589 to 688.7 Apr 2026 Apr 2026 Official leaderboard
12 GPT-5.4 593.2 559.7 to 638 Mar 2026 Mar 2026 Official leaderboard
13 Gemini 3.1 Pro 568.9 514.1 to 633.9 Feb 2026 Feb 2026 Official leaderboard
14 GPT-5.3 Codex 566.5 525.5 to 623.8 Feb 2026 Feb 2026 Official leaderboard
15 GPT-5.2 551.5 515.9 to 596.9 Dec 2025 Dec 2025 Official leaderboard
16 Claude Opus 4.5 533.3 495.7 to 581.8 Nov 2025 Nov 2025 Official leaderboard
17 Gemini 3 Pro 490.5 448.6 to 550.2 Nov 2025 Nov 2025 Official leaderboard
18 GPT-5 478.6 433 to 543.3 Aug 2025 Aug 2025 Official leaderboard
19 GPT-5.1-Codex-Max 477.7 424.8 to 537 Nov 2025 Nov 2025 Official leaderboard
20 o3 417.3 316.6 to 482 Apr 2025 Apr 2025 Official leaderboard
21 Claude Opus 4.1 398.5 331.6 to 486.7 Aug 2025 Aug 2025 Official leaderboard
22 Claude Opus 4 392.6 319.4 to 484.5 May 2025 May 2025 Official leaderboard
23 Claude 3.7 Sonnet 310.0 239.7 to 398.1 Feb 2025 Feb 2025 Official leaderboard
24 o1 237.4 155.9 to 326.3 Dec 2024 Dec 2024 Official leaderboard
25 Claude 3.5 Sonnet (October 2024) 192.1 76.7 to 285.5 Oct 2024 Oct 2024 Official leaderboard
26 o1-preview 140.2 3.9 to 226.8 Sep 2024 Sep 2024 Official leaderboard
27 Claude 3.5 Sonnet (Old) 70.6 ±100.3 Jun 2024 Official leaderboard
28 GPT-4o May 2024 0.0 May 2024 May 2024 Official leaderboard
29 Claude 3 Opus -84.5 -223.1 to 35.1 Feb 2024 Mar 2024 Official leaderboard
30 GPT-4 Turbo Apr 2024 -105.6 -235.6 to -1.2 Apr 2024 Apr 2024 Official leaderboard
#ModelScore ±SettingReleasedRun onSource
1 GPT-6 Astra 859.0 761.1 to 1085.3 Sep 2026 Sep 2026 Official leaderboard
2 Claude Opus 5.5 854.1 765.2 to 1069.3 Sep 2026 Sep 2026 Official leaderboard
3 Claude Fable 5.1 839.1 753.7 to 1019.1 Sep 2026 Sep 2026 Official leaderboard
4 Claude Opus 5 796.2 735.9 to 910.1 Jul 2026 Jul 2026 Official leaderboard
5 Claude Fable 5 778.7 712.9 to 890.9 Jun 2026 Jun 2026 Official leaderboard
6 GPT-5.6 Sol 738.3 681.6 to 855.8 Jul 2026 Jul 2026 Official leaderboard
7 Claude Opus 4.8 708.6 661.3 to 784.3 May 2026 May 2026 Official leaderboard
8 GPT-5.5 705.7 657.8 to 817.1 Apr 2026 Apr 2026 Official leaderboard
9 Claude Opus 4.6 633.0 Feb 2026 Feb 2026 Official leaderboard
10 Claude Opus 4.7 631.2 589 to 688.7 Apr 2026 Apr 2026 Official leaderboard
11 GPT-5.4 593.2 559.7 to 638 Mar 2026 Mar 2026 Official leaderboard
12 Gemini 3.1 Pro 568.9 514.1 to 633.9 Feb 2026 Feb 2026 Official leaderboard
13 GPT-5.3 Codex 566.5 525.5 to 623.8 Feb 2026 Feb 2026 Official leaderboard
14 GPT-5.2 551.5 515.9 to 596.9 Dec 2025 Dec 2025 Official leaderboard
15 Claude Opus 4.5 533.3 495.7 to 581.8 Nov 2025 Nov 2025 Official leaderboard
16 Gemini 3 Pro 490.5 448.6 to 550.2 Nov 2025 Nov 2025 Official leaderboard
17 GPT-5 478.6 433 to 543.3 Aug 2025 Aug 2025 Official leaderboard
18 GPT-5.1-Codex-Max 477.7 424.8 to 537 Nov 2025 Nov 2025 Official leaderboard
19 o3 417.3 316.6 to 482 Apr 2025 Apr 2025 Official leaderboard
20 Claude Opus 4.1 398.5 331.6 to 486.7 Aug 2025 Aug 2025 Official leaderboard
21 Claude Opus 4 392.6 319.4 to 484.5 May 2025 May 2025 Official leaderboard
22 Claude 3.7 Sonnet 310.0 239.7 to 398.1 Feb 2025 Feb 2025 Official leaderboard
23 o1 237.4 155.9 to 326.3 Dec 2024 Dec 2024 Official leaderboard
24 Claude 3.5 Sonnet (October 2024) 192.1 76.7 to 285.5 Oct 2024 Oct 2024 Official leaderboard
25 o1-preview 140.2 3.9 to 226.8 Sep 2024 Sep 2024 Official leaderboard
26 GPT-4o May 2024 0.0 May 2024 May 2024 Official leaderboard
27 Claude 3 Opus -84.5 -223.1 to 35.1 Feb 2024 Mar 2024 Official leaderboard
28 GPT-4 Turbo Apr 2024 -105.6 -235.6 to -1.2 Apr 2024 Apr 2024 Official leaderboard

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI
Own leaderboard 1 leader
GPT-6 Astra 859.0

Source: Official METR Capability Index leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard
0 250 500 750 1000 2024 2025 2026 0 250 500 750 1000 2024 2025 2026

Source: Official METR Capability Index leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in AI research

All 164 in AI research
  • PaperBench In use Agent replicates ICML 2024 papers from scratch. LLM judge grades against author-written hierarchical rubrics. Qwen 3.8 Max 93%
  • MLS-Bench In use Agent must invent one algorithmic change in an ML research scaffold whose gain transfers across settings. Scored relative to human baselines. Claude Fable 5.1 50.3%
  • PostTrainBench 1.2 In use CLI agent gets a base model, one H100 and a time budget to post-train it for a target benchmark. Scored on normalized gain. Claude Fable 5.1 44.6%
  • MLE-bench In use Kaggle-style ML engineering competitions. Agent trains models and is scored against competition leaderboards. GPT-6.2 Astra 96.1%
  • PMPP-Hard In use Agents write GPU kernels for 69 tasks that must be correct, and on 48 of them also run within 1.25 times the reference time. The score is the area under the solved-tasks curve over a 10-hour budget, out of 690 task-hours. Kimi K3 0.71
  • KernelBench In use Replace PyTorch operators with faster custom GPU kernels. Scored on correctness and speedup. MiniMax-M3 28.8%

Data audit

All 10 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (30 entries).
  • Pass Models scored 28 models we track, plus 2 other entries on its leaderboard.
  • Pass Newest score The newest is from September 2026, for Claude Opus 5.5.
  • Pass Frontier models 3 of the top 10 have a score, led by GPT-6 Astra (#4).
  • Note Headroom Scored as an index, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on metr.org: 28 rows for models we track and 2 other entries at the last read.
  • Pass Last read November 21st, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check