Vals RSI Index
Four autonomous model-development tasks: compression, language-model training, harness engineering and post-training.
Autonomous researchVals AI23 models run by Vals AIUpdated 29 September 2026
How it is run
Version 1.1 normalizes each task to frozen baseline, reference and theoretical-maximum anchors; the overall index averages the four task scores.
What it cannot tell you
The percentage is a reference-normalized index, not task accuracy. Task budgets and vendor-native agent harnesses differ.
- Leader
- Claude Opus 5.537.31% · Vals AI's own list
- Top-ten spread
- 13.43%Leader minus tenth place, or last place in smaller fields.
- Median
- 21.76%of the 23 rows published
- No published result
- 26of 46 models on this site
Independent leaderboard · Vals AI
Top 23 of 23 runs by Vals AI, one harness.
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 5.5 | 37.31% |
| 2 | Claude Fable 5.1 | 36.09% |
| 3 | Claude Opus 5 | 33.02% |
| 4 | Gemini 4 Argon | 30.55% |
| 5 | GPT-6 Sol | 28.11% |
| 6 | GPT-6 Astra | 27.07% |
| 7 | Grok 4.6 | 25.46% |
| 8 | Claude Fable 5 | 24.32% |
| 9 | Claude Opus 4.8 | 23.98% |
| 10 | GPT-5.6 Sol | 23.88% |
| 11 | GLM 5.3 | 22.91% |
| 12 | Qwen3.8 Max | 21.76% |
| 13 | Kimi K3 | 20.88% |
| 14 | Grok 4.7 | 20.20% |
| 15 | Gemini 3.8 Flash | 19.93% |
| 16 | Muse Spark 1.3 Max | 19.64% |
| 17 | Claude Opus 4.7 | 19.08% |
| 18 | Gemini 3.7 Flash | 18.27% |
| 19 | GPT-5.5 | 16.48% |
| 20 | Gpt 5.2 2025 12 11 | 15.80% |
| 21 | GPT-5.4 | 15.48% |
| 22 | Claude Opus 4 5 20251101 | 14.82% |
| 23 | Kimi K2.5 Thinking | 6.85% |
No published result tracked
26 of the 46 models on this site have no published number on this benchmark, so their cells in the home table stay blank rather than reading as zero.
Step 5 PreviewGPT-6 LunaGPT-5.6 TerraGLM 5.3 FlashQwen3.8 2.4T A95BQwen3.8-Flash-NextMiMo-V2.6-FlashQwen3.8 Max (0902)Muse Spark 1.2Grok 4.5DeepSeek V4.1 Flash 1017GLM-LegalClaude Haiku 5.5BeamClaude Sonnet 5.5Claude Sonnet 5GPT-6.1 SolGPT-5.6 LunaGemini 3.1 ProDeepSeek V4.1 FlashDeepSeek V4 ProMiniMax M3Hunyuan Hy4 PreviewMiMo V2.6 ProInklingMistral Medium 3.5
Sources
Every page used for this benchmark, read on 17 October 2026.
- Benchmark site or the page that publishes itwww.vals.ai/benchmarks/rsi_index
- Independent resultswww.vals.ai/benchmarks/rsi_index