Benchmark Almanac

MysteryMechanism

GPT-6 Astra 53.1% leads since Sep 2026 Not in model cards

Agents recover a hidden mathematical law from two noisy observations and a few experiments they choose. The submitted expression is run on private probes and passes if its error is under a noise-aware threshold.

Inputs are anonymous, with no domain context. A blinded audit records whether each solved trace named the source domain. GPT-6 Astra led at 53.15% on 2026-10-29.

Made by Vals AI Released Sep 2026 Abstract puzzles and logic Tasks kept private Last model added Oct 2026 Scores refreshed 2 Nov 2026
Level
Expert Specialists with years of training, often over hours of work
Scored
Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers
Agent task Many steps in a terminal, browser, desktop, app, game or robot.
Leaderboard vals.ai
29
Models scored on its own leaderboard

GPT-6 Astra leads with 53.1%

Best score per model from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

29 models. Scroll sideways for the rest.

  1. 53.1 GPT-6 Astra
  2. 49.5 Claude Opus 5.5
  3. 49.1 Claude Sonnet 5.5
  4. 47.8 Claude Fable 5.1
  5. 46.4 GPT-6.1 Sol
  6. 45.5 Gemini 4 Argon
  7. 37.4 Claude Opus 5
  8. 36.5 Gemini 3.8 Flash
  9. 36.0 Muse Spark 1.3
  10. 35.1 Claude Haiku 5.5
  11. 35.1 Muse Wonder Full
  12. 33.3 GPT-5.6 Sol
  13. 30.6 Grok 4.6
  14. 30.2 GPT-6 Sol
  15. 27.5 Muse Wonder
  16. 26.1 Kimi K3.1 Pro
  17. 25.2 Grok 4.7
  18. 23.9 Qwen 3.8 Max
  19. 23.0 GLM-5.3
  20. 22.5 Ember-1
  21. 22.1 Beam 1.0
  22. 21.6 MiMo-V2.6 Flash
  23. 21.2 DeepSeek V4.1 Flash
  24. 20.7 Gemini 3.1 Pro
  25. 19.4 GPT-6 Luna
  26. 18.5 Step 5 Preview
  27. 15.3 MiMo-V2.6 Pro
  28. 14.4 GPT-5.6 Luna
  29. 12.2 Mistral Large 4
  1. 1 GPT-6 Astra 53.1%
  2. 2 Claude Opus 5.5 49.5%
  3. 3 Claude Sonnet 5.5 49.1%
  4. 4 Claude Fable 5.1 47.8%
  5. 5 GPT-6.1 Sol 46.4%
  6. 6 Gemini 4 Argon 45.5%
  7. 7 Claude Opus 5 37.4%
  8. 8 Gemini 3.8 Flash 36.5%
  9. 9 Muse Spark 1.3 max 36%
  10. 10 Claude Haiku 5.5 35.1%
  11. 11 Muse Wonder Full 35.1%
  12. 12 GPT-5.6 Sol 33.3%
  13. 13 Grok 4.6 30.6%
  14. 14 GPT-6 Sol 30.2%
  15. 15 Muse Wonder 27.5%
  16. 16 Kimi K3.1 Pro 26.1%
  17. 17 Grok 4.7 25.2%
  18. 18 Qwen 3.8 Max 23.9%
  19. 19 GLM-5.3 23%
  20. 20 Ember-1 22.5%
  21. 21 Beam 1.0 22.1%
  22. 22 MiMo-V2.6 Flash 21.6%
  23. 23 DeepSeek V4.1 Flash 21.2%
  24. 24 Gemini 3.1 Pro 20.7%
  25. 25 GPT-6 Luna 19.4%
  26. 26 Step 5 Preview 18.5%
  27. 27 MiMo-V2.6 Pro 15.3%
  28. 28 GPT-5.6 Luna 14.4%
  29. 29 Mistral Large 4 12.2%
#ModelScoreSettingDateSource
1 GPT-6 Astra 53.1% Sep 2026 Official leaderboard
2 Claude Opus 5.5 49.5% Sep 2026 Official leaderboard
3 Claude Sonnet 5.5 49.1% Sep 2026 Official leaderboard
4 Claude Fable 5.1 47.8% Sep 2026 Official leaderboard
5 GPT-6.1 Sol 46.4% Sep 2026 Official leaderboard
6 Gemini 4 Argon 45.5% Sep 2026 Official leaderboard
7 Claude Opus 5 37.4% Jul 2026 Official leaderboard
8 Gemini 3.8 Flash 36.5% Sep 2026 Official leaderboard
9 Muse Spark 1.3 36% max Sep 2026 Official leaderboard
10 Claude Haiku 5.5 35.1% Oct 2026 Official leaderboard
11 Muse Wonder Full 35.1% Oct 2026 Official leaderboard
12 GPT-5.6 Sol 33.3% Jul 2026 Official leaderboard
13 Grok 4.6 30.6% Aug 2026 Official leaderboard
14 GPT-6 Sol 30.2% Sep 2026 Official leaderboard
15 Muse Wonder 27.5% Oct 2026 Official leaderboard
16 Kimi K3.1 Pro 26.1% Oct 2026 Official leaderboard
17 Grok 4.7 25.2% Sep 2026 Official leaderboard
18 Qwen 3.8 Max 23.9% Aug 2026 Official leaderboard
19 GLM-5.3 23% Aug 2026 Official leaderboard
20 Ember-1 22.5% Sep 2026 Official leaderboard
21 Beam 1.0 22.1% Oct 2026 Official leaderboard
22 MiMo-V2.6 Flash 21.6% Sep 2026 Official leaderboard
23 DeepSeek V4.1 Flash 21.2% Sep 2026 Official leaderboard
24 Gemini 3.1 Pro 20.7% Feb 2026 Official leaderboard
25 GPT-6 Luna 19.4% Sep 2026 Official leaderboard
26 Step 5 Preview 18.5% Sep 2026 Official leaderboard
27 MiMo-V2.6 Pro 15.3% Sep 2026 Official leaderboard
28 GPT-5.6 Luna 14.4% Jul 2026 Official leaderboard
29 Mistral Large 4 12.2% Oct 2026 Official leaderboard

The best score, 53.1%, costs $1.56 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

OpenAI Xiaomi Corp Cost frontier One model at each effort level
0% 20% 40% 60% $0.01 $0.10 $1.00 $10 Cost per task, log scale GPT-6 Astra GPT-6.1 Sol MiMo-V2.6 Flash GPT-6 Luna 0% 20% 40% 60% $0.01 $0.10 $1.00 $10 Cost per task, log scale GPT-6 Astra GPT-6.1 Sol
ModelEffortHarnessScoreCost per taskOutput tokens
GPT-6 Astra 53.1% $1.56
Claude Opus 5.5 49.5% $4.62
Claude Sonnet 5.5 49.1% $3.45
Claude Fable 5.1 47.8% $5.63
GPT-6.1 Sol 46.4% $0.28
Gemini 4 Argon 45.5% $4.59
Claude Opus 5 37.4% $3.28
Gemini 3.8 Flash 36.5% $1.73
Muse Spark 1.3 max 36% $0.85
Claude Haiku 5.5 35.1% $0.54
Muse Wonder Full 35.1% $4.16
GPT-5.6 Sol 33.3% $0.96
Grok 4.6 30.6% $1.31
GPT-6 Sol 30.2% $0.46
Muse Wonder 27.5% $3.39
Kimi K3.1 Pro 26.1% $0.62
Grok 4.7 25.2% $1.84
Qwen 3.8 Max 23.9% $1.31
GLM-5.3 23% $1.53
Ember-1 22.5% $1.29
Beam 1.0 22.1%
MiMo-V2.6 Flash 21.6% $0.06
DeepSeek V4.1 Flash 21.2% $0.16
Gemini 3.1 Pro 20.7% $1.07
GPT-6 Luna 19.4% $0.03
Step 5 Preview 18.5% $0.60
MiMo-V2.6 Pro 15.3% $0.15
GPT-5.6 Luna 14.4% $0.11
Mistral Large 4 12.2% $1.06

2 models have held number one since its release in Sep 2026

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

Anthropic OpenAI
Own leaderboard 2 leaders
GPT-6 Astra 53.1%

Scores over time

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

On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) GPT-6 Astra Claude Fable 5.1 Claude Opus 5 Gemini 3.1 Pro 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) GPT-6 Astra Claude Opus 5

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

More in Science and reasoning

All 88 in Science and reasoning
  • GPQA Diamond In use 198 graduate-level 'Google-proof' multiple-choice questions in biology, chemistry and physics. Gemini 4 Argon 96.4%
  • Terminal-Bench-Science In use 70 scientific research-workflow tasks across five science areas, run in Harbor containers. Hidden tests give the resolution rate. GPT-6 Astra 68.1%
  • FrontierScience In use Models solve physics, chemistry, and biology problems in an Olympiad track graded by short answer and a Research track graded by 10-point rubrics. Intern-S2-397B 87%
  • BioMysteryBench In use Bioinformatics problems on anonymized real datasets that require real analysis with tools. Rubric-graded, with human-solvable and human-difficult splits. Claude Opus 5 90.1%
  • HiPhO In use Models solve theory problems from 2024 and 2025 physics olympiad exams, some with diagrams. Official marking schemes grade answers and steps, mapped to medal levels. Gemini 3 Pro 78.3%
  • LABBench2 In use 11 biology research sub-tasks (literature, figures, tables, patents, trials and more) with tool access. Macro-averaged accuracy. GPT-5.2 60.9%