FrontierML (Synthetic Sciences)
Agents conduct 25 empirical machine learning projects under fixed budgets. Each task has a verifier and reference scores. Harbor includes the 15 public projects.
On November 19, 2026, Claude Fable 5.2 led at 0.523 mean normalized score. Claude Opus 5.5 followed at 0.399 and GPT-6 Astra at 0.371.
- Made by
- Synthetic Sciences
- Released
- October 2026
- Kind
- AI research tasks
- Level
- Research frontier
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- Public, plus a private test set
- Measures
- Capability
- Language
- English
- Interaction
- Agentic
- Environment
- Terminal, APIs and tools
- Refresh
- One fixed set
- Run by
- A host
- In views
- Agents
- Strongest model tested
- Claude Fable 5.2 #2 on the capability index
- Last model added
- November 2026
- Scores refreshed
- November 20th, 2026
Claude Fable 5.2 leads with 0.52
Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best.
- 1 Claude Fable 5.2 0.52
- 2 Claude Opus 5.5 0.40
- 3 GPT-6 Astra 0.37
- 4
Gemini 4 Carbon 0.35
Source: Official FrontierML (Synthetic Sciences) leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Fable 5.2 | 0.52 | Nov 2026 | Official leaderboard | |
| 2 | Claude Opus 5.5 | 0.40 | Sep 2026 | Official leaderboard | |
| 3 | GPT-6 Astra | 0.37 | Sep 2026 | Official leaderboard | |
| 4 | | 0.35 | Oct 2026 | Official leaderboard |
Rankings by part
We read 15 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Pokémon Battles, Rule Worlds and Watercolour. Claude Fable 5.2 leads 10 of them, and 3 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Pokémon Battles
4 models we track have a score here.
- 1 Claude Fable 5.2 0.75
- 2 Claude Opus 5.5 0.41
- 3 GPT-6 Astra 0.17
- 4
Gemini 4 Carbon 0.16
Rule Worlds
4 models we track have a score here.
- 1 Claude Fable 5.2 0.57
- 2 Claude Opus 5.5 0.53
- 3 GPT-6 Astra 0.47
- 4
Gemini 4 Carbon 0.31
Watercolour
4 models we track have a score here.
- 1 Claude Fable 5.2 0.72
- 2
Gemini 4 Carbon ↑2 0.18
- 3 GPT-6 Astra 0.15
- 4 Claude Opus 5.5 ↓2 0.12
Robot Reward
4 models we track have a score here.
- 1 Claude Fable 5.2 0.72
- 2 GPT-6 Astra ↑1 0.67
- 3
Gemini 4 Carbon ↑1 0.58
- 4 Claude Opus 5.5 ↓2 0.49
Off Book
4 models we track have a score here.
- 1 Claude Fable 5.2 0.65
- 2 GPT-6 Astra ↑1 0.16
- 3 Claude Opus 5.5 ↓1 0.12
- 4
Gemini 4 Carbon 0.12
Rare Successes
4 models we track have a score here.
- 1 GPT-6 Astra ↑2 0.43
- 2 Claude Fable 5.2 ↓1 0.42
- 3
Gemini 4 Carbon ↑1 0.31
- 4 Claude Opus 5.5 ↓2 0.30
Credit Assignment
4 models we track have a score here.
- 1
Gemini 4 Carbon ↑3 0.66
- 2 Claude Opus 5.5 0.57
- 3 Claude Fable 5.2 ↓2 0.30
- 4 GPT-6 Astra ↓1 -0.30
NanoJEV
4 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 0.53
- 2 Claude Fable 5.2 ↓1 0.36
- 3
Gemini 4 Carbon ↑1 0.36
- 4 GPT-6 Astra ↓1 0.31
No Way Back
4 models we track have a score here.
- 1 Claude Fable 5.2 0.64
- 2 Claude Opus 5.5 0.42
- 3 GPT-6 Astra 0.29
- 4
Gemini 4 Carbon 0.10
What the User Did
4 models we track have a score here.
- 1 Claude Fable 5.2 0.52
- 2 GPT-6 Astra ↑1 0.34
- 3
Gemini 4 Carbon ↑1 0.28
- 4 Claude Opus 5.5 ↓2 0.27
Trusted Reader
4 models we track have a score here.
- 1 Claude Fable 5.2 0.62
- 2
Gemini 4 Carbon ↑2 0.60
- 3 Claude Opus 5.5 ↓1 0.50
- 4 GPT-6 Astra ↓1 0.36
Forward
4 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 0.62
- 2 Claude Fable 5.2 ↓1 0.62
- 3
Gemini 4 Carbon ↑1 0.55
- 4 GPT-6 Astra ↓1 0.50
New Rooms
4 models we track have a score here.
- 1 GPT-6 Astra ↑2 0.53
- 2 Claude Fable 5.2 ↓1 0.46
- 3 Claude Opus 5.5 ↓1 0.43
- 4
Gemini 4 Carbon 0.33
Emotive Voice
4 models we track have a score here.
- 1 Claude Fable 5.2 0.63
- 2 Claude Opus 5.5 0.58
- 3
Gemini 4 Carbon ↑1 0.58
- 4 GPT-6 Astra ↓1 0.54
pass@32 = 0
4 models we track have a score here.
- 1 Claude Fable 5.2 0.29
- 2 Claude Opus 5.5 0.25
- 3
Gemini 4 Carbon ↑1 0.07
- 4 GPT-6 Astra ↓1 0.00
2 models have held number one on the official leaderboard since its release in October 2026
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official FrontierML (Synthetic Sciences) leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official FrontierML (Synthetic Sciences) 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
- 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 9 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access Public tasks, plus a private test set.
- Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
- Pass Independent runs Scores from the official leaderboard (4 entries).
- Note Models scored Only 4 models we track.
- Pass Newest score The newest is from November 2026, for Claude Fable 5.2.
- Pass Frontier models 4 of the top 10 have a score, led by Claude Fable 5.2 (#2).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on frontierml.ai: 4 rows for models we track at the last read.
- Pass Last read November 20th, 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.