MysteryMechanism
The official leaderboard hasn’t scored a model released after Gemini 4 Neon (November 2026). Labs report their own numbers for one newer model: Inkling 2 Mini at 23%.
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.
- Made by
- Vals AI
- Released
- September 2026
- Kind
- Abstract puzzles and logic
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- Kept private
- Measures
- Capability
- Subject
- Mathematics
- Language
- English
- Interaction
- Agentic
- Environment
- Terminal
- Refresh
- One fixed set
- Run by
- A host
- In views
- Agents
- Strongest model tested
- GPT-6.2 Astra #1 on the capability index
- Last in a system card
- November 2026
- Last model added
- November 2026
- Scores refreshed
- November 19th, 2026
2 models share the top score of 53.1%
Best score per model from the benchmark's own leaderboard, read on November 19th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.
GPT-6 Astra leads with 53.1%, claims included
Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model. Striped bars are claims, and dotted bars are reruns by Epoch AI or another board.
- 53.1 GPT-6 Astra
- 53.1 GPT-6.2 Astra
- 51.8 Claude Fable 5.2
- 49.5 Claude Opus 5.5
- 49.1 Claude Sonnet 5.5
- 47.8 Claude Fable 5.1
- 46.4 GPT-6.1 Sol
- 45.5
Gemini 4 Argon
- 45.0
Gemini 4 Neon
- 37.4 Claude Opus 5
- 36.9 GPT-6.2 Sol
- 36.5
Gemini 3.8 Flash
- 36.0
Muse Spark 1.3
- 35.1 Claude Haiku 5.5
- 35.1
Muse Wonder Full
- 33.3 GPT-5.6 Sol
- 30.6 Grok 4.6
- 30.2 GPT-6 Sol
- 27.5
Muse Wonder
- 26.1 Kimi K3.1 Pro
- 25.2 Grok 4.7
- 25.2
Inkling 2 Small
- 23.9
Qwen 3.8 Max
- 23.0 GLM-5.3
- 22.5
Ember-1
- 22.1 Beam 1.0
- 21.6 MiMo-V2.6 Flash
- 21.2
DeepSeek V4.1 Flash
- 20.7
Gemini 3.1 Pro
- 19.4 GPT-6 Luna
- 18.5
Step 5 Preview
- 15.3 MiMo-V2.6 Pro
- 14.4 GPT-5.6 Luna
- 12.2
Mistral Large 4 Preview
- 1 GPT-6 Astra 53.1%
- 2 GPT-6.2 Astra 53.1%
- 3 Claude Fable 5.2 51.8%
- 4 Claude Opus 5.5 49.5%
- 5 Claude Sonnet 5.5 49.1%
- 6 Claude Fable 5.1 47.8%
- 7 GPT-6.1 Sol 46.4%
- 8
Gemini 4 Argon 45.5%
- 9
Gemini 4 Neon 45%
- 10 Claude Opus 5 37.4%
- 11 GPT-6.2 Sol 36.9%
- 12
Gemini 3.8 Flash 36.5%
- 13
Muse Spark 1.3 max 36%
- 14 Claude Haiku 5.5 35.1%
- 15
Muse Wonder Full 35.1%
- 16 GPT-5.6 Sol 33.3%
- 17 Grok 4.6 30.6%
- 18 GPT-6 Sol 30.2%
- 19
Muse Wonder 27.5%
- 20 Kimi K3.1 Pro 26.1%
- 21 Grok 4.7 25.2%
- 22
Inkling 2 Small 25.2%
- 23
Qwen 3.8 Max 23.9%
- 24 GLM-5.3 23%
- 25
Ember-1 22.5%
- 26 Beam 1.0 22.1%
- 27 MiMo-V2.6 Flash 21.6%
- 28
DeepSeek V4.1 Flash 21.2%
- 29
Gemini 3.1 Pro 20.7%
- 30 GPT-6 Luna 19.4%
- 31
Step 5 Preview 18.5%
- 32 MiMo-V2.6 Pro 15.3%
- 33 GPT-5.6 Luna 14.4%
- 34
Mistral Large 4 Preview 12.2%
Source: Official MysteryMechanism leaderboard
- 53.1 GPT-6 Astra
- 53.1 GPT-6.2 Astra
- 51.8 Claude Fable 5.2
- 49.5 Claude Opus 5.5
- 49.1 Claude Sonnet 5.5
- 47.8 Claude Fable 5.1
- 46.4 GPT-6.1 Sol
- 45.5
Gemini 4 Argon
- 45.0
Gemini 4 Neon
- 37.4 Claude Opus 5
- 36.9 GPT-6.2 Sol
- 36.5
Gemini 3.8 Flash
- 36.0
Muse Spark 1.3
- 35.1 Claude Haiku 5.5
- 35.1
Muse Wonder Full
- 33.3 GPT-5.6 Sol
- 30.6 Grok 4.6
- 30.2 GPT-6 Sol
- 27.5
Muse Wonder
- 26.1 Kimi K3.1 Pro
- 25.2 Grok 4.7
- 25.2
Inkling 2 Small
- 23.9
Qwen 3.8 Max
- 23.0
Inkling 2 Mini
- 23.0 GLM-5.3
- 22.5
Ember-1
- 22.1 Beam 1.0
- 21.6 MiMo-V2.6 Flash
- 21.2
DeepSeek V4.1 Flash
- 20.7
Gemini 3.1 Pro
- 19.4 GPT-6 Luna
- 18.5
Step 5 Preview
- 15.3 MiMo-V2.6 Pro
- 14.4 GPT-5.6 Luna
- 12.2
Mistral Large 4 Preview
- 1 GPT-6 Astra 53.1%
- 2 GPT-6.2 Astra 53.1%
- 3 Claude Fable 5.2 51.8%
- 4 Claude Opus 5.5 49.5%
- 5 Claude Sonnet 5.5 49.1%
- 6 Claude Fable 5.1 47.8%
- 7 GPT-6.1 Sol 46.4%
- 8
Gemini 4 Argon 45.5%
- 9
Gemini 4 Neon 45%
- 10 Claude Opus 5 37.4%
- 11 GPT-6.2 Sol 36.9%
- 12
Gemini 3.8 Flash 36.5%
- 13
Muse Spark 1.3 max 36%
- 14 Claude Haiku 5.5 35.1%
- 15
Muse Wonder Full 35.1%
- 16 GPT-5.6 Sol 33.3%
- 17 Grok 4.6 30.6%
- 18 GPT-6 Sol 30.2%
- 19
Muse Wonder 27.5%
- 20 Kimi K3.1 Pro 26.1%
- 21 Grok 4.7 25.2%
- 22
Inkling 2 Small 25.2%
- 23
Qwen 3.8 Max 23.9%
- 24
Inkling 2 Mini 23%
- 25 GLM-5.3 23%
- 26
Ember-1 22.5%
- 27 Beam 1.0 22.1%
- 28 MiMo-V2.6 Flash 21.6%
- 29
DeepSeek V4.1 Flash 21.2%
- 30
Gemini 3.1 Pro 20.7%
- 31 GPT-6 Luna 19.4%
- 32
Step 5 Preview 18.5%
- 33 MiMo-V2.6 Pro 15.3%
- 34 GPT-5.6 Luna 14.4%
- 35
Mistral Large 4 Preview 12.2%
Sources: Official MysteryMechanism leaderboard and labs’ model cards
| # | Model | Score ± 95% CI | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-6 Astra | 53.1% ±6.6 | Sep 2026 | Official leaderboard | |
| 2 | GPT-6.2 Astra | 53.1% ±6.6 | Nov 2026 | Official leaderboard | |
| 3 | Claude Fable 5.2 | 51.8% ±6.6 | Nov 2026 | Official leaderboard | |
| 4 | Claude Opus 5.5 | 49.5% ±6.6 | Sep 2026 | Official leaderboard | |
| 5 | Claude Sonnet 5.5 | 49.1% ±6.6 | Sep 2026 | Official leaderboard | |
| 6 | Claude Fable 5.1 | 47.8% ±6.6 | Sep 2026 | Official leaderboard | |
| 7 | GPT-6.1 Sol | 46.4% ±6.6 | Sep 2026 | Official leaderboard | |
| 8 | | 45.5% ±6.6 | Sep 2026 | Official leaderboard | |
| 9 | | 45% ±6.6 | Nov 2026 | Official leaderboard | |
| 10 | Claude Opus 5 | 37.4% ±6.4 | Jul 2026 | Official leaderboard | |
| 11 | GPT-6.2 Sol | 36.9% ±6.4 | Nov 2026 | Official leaderboard | |
| 12 | | 36.5% ±6.3 | Sep 2026 | Official leaderboard | |
| 13 | | 36% ±6.3 | max | Sep 2026 | Official leaderboard |
| 14 | Claude Haiku 5.5 | 35.1% ±6.3 | Oct 2026 | Official leaderboard | |
| 15 | | 35.1% ±6.3 | Oct 2026 | Official leaderboard | |
| 16 | GPT-5.6 Sol | 33.3% ±6.2 | Jul 2026 | Official leaderboard | |
| 17 | Grok 4.6 | 30.6% ±6.1 | Aug 2026 | Official leaderboard | |
| 18 | GPT-6 Sol | 30.2% ±6.1 | Sep 2026 | Official leaderboard | |
| 19 | | 27.5% ±5.9 | Oct 2026 | Official leaderboard | |
| 20 | Kimi K3.1 Pro | 26.1% ±5.8 | Oct 2026 | Official leaderboard | |
| 21 | Grok 4.7 | 25.2% ±5.7 | Sep 2026 | Official leaderboard | |
| 22 | | 25.2% ±5.7 | Nov 2026 | Official leaderboard | |
| 23 | | 23.9% ±5.6 | Aug 2026 | Official leaderboard | |
| 24 | GLM-5.3 | 23% ±5.5 | Aug 2026 | Official leaderboard | |
| 25 | | 22.5% ±5.5 | Sep 2026 | Official leaderboard | |
| 26 | Beam 1.0 | 22.1% ±5.5 | Oct 2026 | Official leaderboard | |
| 27 | MiMo-V2.6 Flash | 21.6% ±5.4 | Sep 2026 | Official leaderboard | |
| 28 | | 21.2% ±5.4 | Sep 2026 | Official leaderboard | |
| 29 | | 20.7% ±5.3 | Feb 2026 | Official leaderboard | |
| 30 | GPT-6 Luna | 19.4% ±5.2 | Sep 2026 | Official leaderboard | |
| 31 | | 18.5% ±5.1 | Sep 2026 | Official leaderboard | |
| 32 | MiMo-V2.6 Pro | 15.3% ±4.7 | Sep 2026 | Official leaderboard | |
| 33 | GPT-5.6 Luna | 14.4% ±4.6 | Jul 2026 | Official leaderboard | |
| 34 | | 12.2% ±4.3 | Oct 2026 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-6 Astra | 53.1% | Sep 2026 | Official leaderboard | |
| 2 | GPT-6.2 Astra | 53.1% | Nov 2026 | Official leaderboard | |
| 3 | Claude Fable 5.2 | 51.8% | Nov 2026 | Official leaderboard | |
| 4 | Claude Opus 5.5 | 49.5% | Sep 2026 | Official leaderboard | |
| 5 | Claude Sonnet 5.5 | 49.1% | Sep 2026 | Official leaderboard | |
| 6 | Claude Fable 5.1 | 47.8% | Sep 2026 | Official leaderboard | |
| 7 | GPT-6.1 Sol | 46.4% | Sep 2026 | Official leaderboard | |
| 8 | 45.5% | Sep 2026 | Official leaderboard | ||
| 9 | 45% | Nov 2026 | Official leaderboard | ||
| 10 | Claude Opus 5 | 37.4% | Jul 2026 | Official leaderboard | |
| 11 | GPT-6.2 Sol | 36.9% | Nov 2026 | Official leaderboard | |
| 12 | 36.5% | Sep 2026 | Official leaderboard | ||
| 13 | 36% | max | Sep 2026 | Official leaderboard | |
| 14 | Claude Haiku 5.5 | 35.1% | Oct 2026 | Official leaderboard | |
| 15 | 35.1% | Oct 2026 | Official leaderboard | ||
| 16 | GPT-5.6 Sol | 33.3% | Jul 2026 | Official leaderboard | |
| 17 | Grok 4.6 | 30.6% | Aug 2026 | Official leaderboard | |
| 18 | GPT-6 Sol | 30.2% | Sep 2026 | Official leaderboard | |
| 19 | 27.5% | Oct 2026 | Official leaderboard | ||
| 20 | Kimi K3.1 Pro | 26.1% | Oct 2026 | Official leaderboard | |
| 21 | Grok 4.7 | 25.2% | Sep 2026 | Official leaderboard | |
| 22 | 25.2% | Nov 2026 | Official leaderboard | ||
| 23 | 23.9% | Aug 2026 | Official leaderboard | ||
| 24 | 23% | Nov 2026 | Lab model card | ||
| 25 | GLM-5.3 | 23% | Aug 2026 | Official leaderboard | |
| 26 | 22.5% | Sep 2026 | Official leaderboard | ||
| 27 | Beam 1.0 | 22.1% | Oct 2026 | Official leaderboard | |
| 28 | MiMo-V2.6 Flash | 21.6% | Sep 2026 | Official leaderboard | |
| 29 | 21.2% | Sep 2026 | Official leaderboard | ||
| 30 | 20.7% | Feb 2026 | Official leaderboard | ||
| 31 | GPT-6 Luna | 19.4% | Sep 2026 | Official leaderboard | |
| 32 | 18.5% | Sep 2026 | Official leaderboard | ||
| 33 | MiMo-V2.6 Pro | 15.3% | Sep 2026 | Official leaderboard | |
| 34 | GPT-5.6 Luna | 14.4% | Jul 2026 | Official leaderboard | |
| 35 | 12.2% | Oct 2026 | Official leaderboard |
Labs report scores for 1 model in their cards. Switch the scores above to rank them together.
Accuracy against latency
The Vals scatter chart with its x axis switched from Cost to Latency: mean time per test against accuracy, on a log scale. The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.
| Model | Setting | Accuracy | Latency (mean per test) |
|---|---|---|---|
| GPT-6 Astra | 53.2% | 488.6 s | |
| GPT-6.2 Astra | 53.2% | 634.08 s | |
| Claude Fable 5.2 | 51.8% | 292.06 s | |
| Claude Opus 5.5 | 49.6% | 1,179 s | |
| Claude Sonnet 5.5 | 49.1% | 1,410 s | |
| Claude Fable 5.1 | 47.8% | 1,061 s | |
| GPT-6.1 Sol | 46.4% | 437.02 s | |
| 45.5% | 1,495 s | ||
| 45.1% | 1,893 s | ||
| Claude Opus 5 | 37.4% | 1,013 s | |
| GPT-6.2 Sol | 36.9% | 654.39 s | |
| 36.5% | 322.24 s | ||
| max | 36% | 451.8 s | |
| Claude Haiku 5.5 | 35.1% | 805.03 s | |
| 35.1% | 2,494 s | ||
| GPT-5.6 Sol | 33.3% | 522.74 s | |
| Grok 4.6 | 30.6% | 1,675 s | |
| GPT-6 Sol | 30.2% | 373.43 s | |
| 27.5% | 1,097 s | ||
| Kimi K3.1 Pro | 26.1% | 359.9 s | |
| 25.2% | 1,017 s | ||
| Grok 4.7 | 25.2% | 1,053 s | |
| 23.9% | 2,269 s | ||
| GLM-5.3 | 23% | 1,842 s | |
| 22.5% | 1,215 s | ||
| Beam 1.0 | 22.1% | 144.31 s | |
| MiMo-V2.6 Flash | 21.6% | 2,108 s | |
| 21.2% | 484.33 s | ||
| 20.7% | 360.46 s | ||
| GPT-6 Luna | 19.4% | 361.21 s | |
| 18.5% | 1,760 s | ||
| MiMo-V2.6 Pro | 15.3% | 2,640 s | |
| GPT-5.6 Luna | 14.4% | 493.05 s | |
| 12.2% | 2,137 s |
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.
Source: Official MysteryMechanism leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| GPT-6 Astra | 53.1% | $1.56 | 21,201 | ||
| GPT-6.2 Astra | 53.1% | $1.72 | 26,792 | ||
| Claude Fable 5.2 | 51.8% | $1.55 | 27,779 | ||
| Claude Opus 5.5 | 49.5% | $4.06 | 131,452 | ||
| Claude Sonnet 5.5 | 49.1% | $2.98 | 193,418 | ||
| Claude Fable 5.1 | 47.8% | $5.63 | 80,806 | ||
| GPT-6.1 Sol | 46.4% | $0.28 | 20,186 | ||
| 45.5% | $4.59 | 96,886 | |||
| 45% | $6.63 | 259,853 | |||
| Claude Opus 5 | 37.4% | $3.28 | 81,310 | ||
| GPT-6.2 Sol | 36.9% | $0.49 | 30,074 | ||
| 36.5% | $1.73 | 80,442 | |||
| max | 36% | $0.85 | 91,046 | ||
| Claude Haiku 5.5 | 35.1% | $0.54 | 186,893 | ||
| 35.1% | $4.16 | 99,500 | |||
| GPT-5.6 Sol | 33.3% | $0.96 | 27,506 | ||
| Grok 4.6 | 30.6% | $1.31 | 107,454 | ||
| GPT-6 Sol | 30.2% | $0.46 | 30,502 | ||
| 27.5% | $3.39 | 102,039 | |||
| Kimi K3.1 Pro | 26.1% | $0.62 | 38,095 | ||
| Grok 4.7 | 25.2% | $1.84 | 80,598 | ||
| 25.2% | $0.16 | 84,377 | |||
| 23.9% | $1.31 | 108,723 | |||
| GLM-5.3 | 23% | $1.53 | 156,748 | ||
| 22.5% | $1.29 | 60,449 | |||
| Beam 1.0 | 22.1% | 36,624 | |||
| MiMo-V2.6 Flash | 21.6% | $0.06 | 115,716 | ||
| 21.2% | $0.16 | ||||
| 20.7% | $1.07 | 36,840 | |||
| GPT-6 Luna | 19.4% | $0.03 | 36,248 | ||
| 18.5% | $0.60 | 126,039 | |||
| MiMo-V2.6 Pro | 15.3% | $0.15 | 111,531 | ||
| GPT-5.6 Luna | 14.4% | $0.11 | 42,721 | ||
| 12.2% | $1.06 | 236,882 |
2 models have held number one on the official leaderboard since its release in September 2026 2 models have held number one with claims included since its release in September 2026
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official MysteryMechanism leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Sources: Official MysteryMechanism leaderboard and labs’ model cards
Labs report it only in system cards
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark. Some system cards do, below.
One system card reports it
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Nov 20, 2026 | | 23 | MysteryMechanism | Capabilities |
More in Science and reasoning
- GPQA Diamond In use 198 graduate-level 'Google-proof' multiple-choice questions in biology, chemistry and physics.
Gemini 4 Carbon 97%
- 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%
- 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 Fable 5.2 83.7%
- 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%
- 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%
Data audit
All 11 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access The tasks are private.
- Pass Confidence Checked against a primary source.
Scores 5 pass
- Pass Independent runs Scores from the official leaderboard (34 entries) and system cards (1 score).
- Pass Models scored 35 models we track.
- Pass Newest score The newest is from November 2026, for Inkling 2 Mini.
- Pass Frontier models 8 of the top 10 have a score, led by GPT-6.2 Astra (#1).
- Pass Headroom The best is 53.1%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on vals.ai: 34 rows for models we track at the last read.
- Pass Last read November 19th, 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.