Flex-RSI
Agents improve a starter solution and their own tools within a fixed budget. A hidden test split measures the final solution.
The leaderboard ranks models by average placement across the tasks each model ran. Models need five tasks for an overall rank.
- Made by
- Flex-RSI team
- Released
- October 2026
- Kind
- Machine learning engineering
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Agentic
- Environment
- Code sandbox
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Agents
- Strongest model tested
- GPT-6 Astra #4 on the capability index
- Last model added
- October 2026
- Scores refreshed
- November 20th, 2026
Claude Opus 5.5 leads with 1.52
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 Opus 5.5 1.52
- 2
Gemini 4 Argon 2.00
- 3 GPT-5.6 Sol Unranked (fewer than five tasks) 2.75
- 4 GPT-6.1 Sol 2.76
- 5
Qwen 3.6 35B-A3B Unranked (fewer than five tasks) 3.50
- 6 GPT-6 Astra 3.65
- 7
Qwen3.8-Flash-Next 4.68
- 8
Gemini 3.7 Flash Unranked (fewer than five tasks) 5.50
- 9
Qwen 3.8 27B 5.84
- 10 Claude Sonnet 5.5 6.00
- 11
Gemini 3.8 Flash 6.50
- 12
DeepSeek V4.1 Flash 1017 6.82
- 13
Muse Wonder Unranked (fewer than five tasks) 7.33
- 14 GLM-5.3-Flash 7.40
- 15
Gemma 4 31B IT 7.59
- 16 Grok 4.6 Unranked (fewer than five tasks) 8.00
- 17
Muse Spark 1.3 Unranked (fewer than five tasks) 8.00
- 18
MiniMax-M3 8.77
Source: Official Flex-RSI leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 | 1.52 | Sep 2026 | Official leaderboard | |
| 2 | | 2.00 | Sep 2026 | Official leaderboard | |
| 3 | GPT-5.6 Sol | 2.75 | Unranked (fewer than five tasks) | Jul 2026 | Official leaderboard |
| 4 | GPT-6.1 Sol | 2.76 | Sep 2026 | Official leaderboard | |
| 5 | | 3.50 | Unranked (fewer than five tasks) | Apr 2026 | Official leaderboard |
| 6 | GPT-6 Astra | 3.65 | Sep 2026 | Official leaderboard | |
| 7 | | 4.68 | Aug 2026 | Official leaderboard | |
| 8 | | 5.50 | Unranked (fewer than five tasks) | Aug 2026 | Official leaderboard |
| 9 | | 5.84 | Aug 2026 | Official leaderboard | |
| 10 | Claude Sonnet 5.5 | 6.00 | Sep 2026 | Official leaderboard | |
| 11 | | 6.50 | Sep 2026 | Official leaderboard | |
| 12 | | 6.82 | Oct 2026 | Official leaderboard | |
| 13 | | 7.33 | Unranked (fewer than five tasks) | Oct 2026 | Official leaderboard |
| 14 | GLM-5.3-Flash | 7.40 | Aug 2026 | Official leaderboard | |
| 15 | | 7.59 | Apr 2026 | Official leaderboard | |
| 16 | Grok 4.6 | 8.00 | Unranked (fewer than five tasks) | Aug 2026 | Official leaderboard |
| 17 | | 8.00 | Unranked (fewer than five tasks) | Sep 2026 | Official leaderboard |
| 18 | | 8.77 | Jun 2026 | Official leaderboard |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Flex-RSI 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 Flex-RSI 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 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 (18 entries).
- Pass Models scored 18 models we track.
- Pass Newest score The newest is from October 2026, for DeepSeek V4.1 Flash 1017.
- Pass Frontier models 5 of the top 10 have a score, led by GPT-6 Astra (#4).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on flex-rsi.github.io: 18 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.