LisanBench Token Efficiency: Anthropic and Muse Wonder Lead Tokens per Step
scaling01 · x · 2026-10-24
A third slice of scaling01's LisanBench update: measured by tokens per step and tokens per point, Anthropic (Opus 5.5 low at 7.7 tokens/step) and the Muse Wonder family dominate token efficiency, complementing the earlier perf-per-dollar table.
More from Models
- Kimi k3.1's 'ridiculously bad' usage limits overshadow a strong model — haider1 · 2026-10-24
- Users Report Gemini Severely Hallucinating and Unusable for Days — PsychBreacher · 2026-10-24
- Muse's invisible-messages bug: tool-call progress never renders, Meta silent for 36 hours — inductionheads · 2026-10-24
- Duplicate Link Repost of Kimi K3.1 Pro xdev-bench-hard Benchmark Tweet — ricklamers · 2026-10-24
- Kimi K3.1 Pro Jumps to 81% pass@3 on Cyber Exploit Benchmark, Up From K3's 33% — ricklamers · 2026-10-24
- AI-Assisted Paper on Z-Library Cites a TorrentFreak Article That Doesn't Exist — TangerineDream · 2026-10-24