Vals MysteryMechanism (MysteryMechanism)
We show this table for reference; we do not rank on it.
Can agents rediscover sealed mathematical mechanisms through bounded experiments?
Evaluation results published by Vals AI. Published leaderboard aggregates and evaluation settings only. Dataset contents are not included.
MysteryMechanism score on MysteryMechanism — November 11, 2026
We mirror the published mysterymechanism score view for MysteryMechanism. GPT-6 Astra leads the public snapshot at 53.15%, followed by Claude Fable 5.2 (51.80%) and Claude Opus 5.5 (49.55%). We do not use these results to rank models overall.
GPT-6 Astra
OpenAI
max reasoning
Claude Fable 5.2
Anthropic
Claude Opus 5.5
Anthropic
31 modelsAgenticCurrentDisplay onlyUpdated November 11, 2026
MysteryMechanism score table (31 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | GPT-6 AstraOpenAImax reasoning | 53.15% | Not reported | Closed |
| 2 | Claude Fable 5.2Anthropic | 51.80% | Not reported | Closed |
| 3 | Claude Opus 5.5Anthropic | 49.55% | Not reported | Closed |
| 4 | Claude Sonnet 5.5Anthropic | 49.10% | Not reported | Closed |
| 5 | Claude Fable 5.1Anthropic | 47.75% | Not reported | Closed |
| 6 | GPT-6.1 SolOpenAImax reasoning | 46.40% | Not reported | Closed |
| 7 | Gemini 4 ArgonGooglehigh reasoning | 45.49% | Not reported | Closed |
| 8 | Claude Opus 5Anthropic | 37.39% | Not reported | Closed |
| 9 | Gemini 3.8 FlashGooglehigh reasoning | 36.49% | Not reported | Closed |
| 10 | Muse Spark 1.3 MaxMetamax reasoning | 36.04% | Not reported | Unknown |
| 11 | Claude Haiku 5.5Anthropic | 35.13% | Not reported | Closed |
| 12 | Muse Wonder FullMetamax reasoning | 35.13% | Not reported | Closed |
| 13 | GPT-5.6 SolOpenAImax reasoning | 33.33% | Not reported | Closed |
| 14 | Grok 4.6xAIhigh reasoning | 30.63% | Not reported | Closed |
| 15 | GPT-6 SolOpenAImax reasoning | 30.18% | Not reported | Closed |
| 16 | Muse WonderMetamax reasoning | 27.48% | Not reported | Closed |
| 17 | Kimi K3.1 ProMoonshot AImax reasoning | 26.13% | Not reported | Open |
| 18 | Inkling 2 SmallThinking Machines Lab0.99 reasoning | 25.23% | Not reported | Closed |
| 19 | Grok 4.7xAIxhigh reasoning | 25.23% | Not reported | Closed |
| 20 | Qwen3.8 MaxAlibaba | 23.87% | Not reported | Open |
| 21 | GLM-5.3Z.AImax reasoning | 22.97% | Not reported | Open |
| 22 | Ember-1Fireworks | 22.52% | Not reported | Closed |
| 23 | Beam 1.0Reflection | 22.07% | Not reported | Unknown |
| 24 | MiMo-V2.6-FlashXiaomi | 21.62% | Not reported | Open |
| 25 | DeepSeek V4.1 Flash 1017DeepSeekhigh reasoning | 21.17% | Not reported | Open |
| 26 | Gemini 3.1 Pro PreviewGooglehigh reasoning | 20.72% | Not reported | Unknown |
| 27 | GPT-6 LunaOpenAImax reasoning | 19.37% | Not reported | Closed |
| 28 | Step 5 PreviewStepFun | 18.47% | Not reported | Open |
| 29 | MiMo-V2.6-ProXiaomi | 15.31% | Not reported | Open |
| 30 | GPT-5.6 LunaOpenAImax reasoning | 14.41% | Not reported | Closed |
| 31 | Mistral Large 4Mistralhigh reasoning | 12.16% | Not reported | Open |
How MysteryMechanism is shown here
BenchLM mirrors the public Vals AI MysteryMechanism leaderboard captured from https://www.vals.ai/benchmarks/mysterymechanism and updated by Vals on November 11, 2026. The snapshot preserves overall scores, uncertainty, latency, cost-per-test metadata, and task-level scores where Vals publishes them.
MysteryMechanism is display only on BenchLM. Vals proprietary or Vals-hosted aggregate views are useful context, but BenchLM does not use them as weighted ranking inputs or as a replacement for benchmark-native source records.
Snapshot
The published MysteryMechanism snapshot places GPT-6 Astra first at 53.15%. The third row is 3.60 points behind. The broader top-10 range is 17.12 points, so the table still separates the published systems.
31 models have been evaluated on MysteryMechanism. The benchmark falls in the Agentic category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. MysteryMechanism is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About MysteryMechanism
Year
2026
Tasks
Sealed mathematical mechanisms probed under an experiment budget
Format
Accuracy score
Difficulty
Agentic scientific discovery
An agent is given a sealed mechanism it can only probe through a bounded budget of experiments, and is scored on how well it recovers the underlying rule. The task set is private and the runner is external, so the board stays display only and outside weighted rankings.
Freshness and provenance
Version
MysteryMechanism 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does MysteryMechanism measure?
Can agents rediscover sealed mathematical mechanisms through bounded experiments?
Which model leads the published MysteryMechanism snapshot?
GPT-6 Astra currently leads the published MysteryMechanism snapshot with 53.15% mysterymechanism score. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on MysteryMechanism?
The November 11, 2026 snapshot contains 31 AI models.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.