Benchmark Almanac

RSI Bench

GPT-6 Astra 38.9 leads since September 2026
Not in model cards

Agents get a starting environment, a GPU budget and a verifier suite for an AI R&D task such as data curation or post-training. The score is mean reward per task, averaged across tasks.

Announced August 7th, 2026. Tasks run on Harbor with Modal GPUs. Four sample tasks are public. The leaderboard lists 8 models.

Made by
Scale AI
Released
August 2026
Kind
AI research tasks
Level
Research frontier
Scored
Checked automatically
Answers
Agent task
Tasks
Public, plus a private test set
Strongest model tested
GPT-6 Astra #3 on the capability index
Last model added
September 2026
Scores refreshed
November 12th, 2026
8
Models scored on its own leaderboard

GPT-6 Astra leads with 38.9

Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best.

  1. 38.9 GPT-6 Astra
  2. 37.0 GPT-6.1 Sol
  3. 35.2 Claude Opus 5.5
  4. 35.2 Claude Sonnet 5.5
  5. 33.4 Qwen 3.8 Max
  6. 32.8 Kimi K3
  7. 32.3 GLM-5.3
  8. 32.1 DeepSeek V4.1 Flash
  1. 1 GPT-6 Astra max 38.9
  2. 2 GPT-6.1 Sol max 37.0
  3. 3 Claude Opus 5.5 max 35.2
  4. 4 Claude Sonnet 5.5 max 35.2
  5. 5 Qwen 3.8 Max max 33.4
  6. 6 Kimi K3 max 32.8
  7. 7 GLM-5.3 max 32.3
  8. 8 DeepSeek V4.1 Flash max 32.1

Source: Official RSI Bench leaderboard

#ModelScoreSettingDateSource
1 GPT-6 Astra 38.9 max, Codex Sep 2026 Official leaderboard
2 GPT-6.1 Sol 37.0 max, Codex Sep 2026 Official leaderboard
3 Claude Opus 5.5 35.2 max, Claude Code Sep 2026 Official leaderboard
4 Claude Sonnet 5.5 35.2 max, Claude Code Sep 2026 Official leaderboard
5 Qwen 3.8 Max 33.4 max, terminus-2 Aug 2026 Official leaderboard
6 Kimi K3 32.8 max, terminus-2 Jul 2026 Official leaderboard
7 GLM-5.3 32.3 max, terminus-2 Aug 2026 Official leaderboard
8 DeepSeek V4.1 Flash 32.1 max, terminus-2 Sep 2026 Official leaderboard

3 models have held number one on the official leaderboard since its release in August 2026

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

Moonshot AI Qwen OpenAI
Own leaderboard 3 leaders
GPT-6 Astra 38.9

Source: Official RSI Bench leaderboard

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
32.0 34.0 36.0 38.0 40.0 32.0 34.0 36.0 38.0 40.0

Source: Official RSI Bench leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in AI research

All 104 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 Astra 93.8%
  • KernelBench In use Replace PyTorch operators with faster custom GPU kernels. Scored on correctness and speedup. MiniMax-M3 28.8%
  • 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. Inkling 2 Small 494.3