New benchmark's outlier forces a custom x-axis as Claude Opus 5.5 tops test-time scaling
andrew_n_carr · x · 2026-10-27
Andrew Carr shared a FrontierCS test-time scaling chart plotting FCI against per-run cost budget, where Claude Opus 5.5 is such an outlier it required a custom x-axis. DeepSeek V4.1 Flash hits 30 FCI at just $0.2, while Claude Opus 5.5, GPT-6 Astra and Gemini 4 Argon lead the high-budget tier, with GLM 5.3, Kimi K3 and Qwen 3.6 Max in mid ranges.
More from Models
- Opus 3's origin story survives repeated context compactions, seam after seam — repligate · 2026-10-27
- Human-written essay flagged as Mixed by AI detector Pangram in blogger's test — Aizkmusic · 2026-10-27
- AI community buzzes: Artificial Analysis hints a major release is imminent — rand_longevity · 2026-10-27
- Cryptographer: No significant cryptanalysis result from frontier AI labs since late July — matthew_d_green · 2026-10-27
- Kimi Team proposes Attention Residuals, replacing fixed residual accumulation with learned attention — burny_tech · 2026-10-27
- Two open checkpoints show a full JEPA-pretrain-to-GRPO RL recipe on small LLMs — burny_tech · 2026-10-27