FrontierML benchmark shows open models still far behind closed ones on end-to-end ML work
scaling01 · x · 2026-10-27
FrontierML, released just two weeks ago by inductionlabs, is likely untainted by lab-specific tuning — and its first results show a large gap between open and closed models on complex, end-to-end ML tasks.
Key points from the chart:
- Claude Opus 5.5 leads closed models at 59, ahead of Sonnet 5.5 (50) and GPT-6 Astra (44); human researchers still top out above 60
- Best open models, Kimi K3 (20) and Kimi K3.1 Pro (29), trail far behind; DeepSeek V4.1 Flash and GLM 5.3 sit below 25, Beam 1.0 under 10
- X-axis is total spend per task (wallet spend + agent API cost at $10/$50/$200 tiers), showing closed models' high scores come at higher cost
Related event: FrontierML: Kimi K3.1 Pro Tops Open Models but Lags Closed Frontier(4 posts)→
More from Models
- Nat Lambert: Cohere's model is narrow RL training on internal infra ill-suited for open weights — natolambert · 2026-10-27
- OpenAI teases DOTS is coming to Europe 'soon' — mark_k · 2026-10-27
- Clifford Pickover challenges AI to decipher a message he encoded in Norse runes — pickover · 2026-10-27
- ChatGPT's crawler sample code strips out all Schema markup — VeryWellVersed · 2026-10-27
- Cohere sandbox: Qwen model tried to upload probe file to Beijing cloud server unprompted — simonguozirui · 2026-10-27
- Kimi K3.1 Pro tops open-weight models on FrontierML benchmark, still far behind Opus 5.5 — teortaxesTex · 2026-10-27