METR Extends Time-Horizon Benchmark: Opus 5.5 Hits 80% Success at ~20-Hour Tasks
ben_j_todd · x · 2026-11-21
METR has extended its original time-horizon benchmark and measured Opus 5.5: as shown in Figure 28, the task duration at an 80% success rate is roughly 20 hours, essentially in line with the 21 hours indicated in earlier charts.
More from Models
- GPT-6.2 Astra ranks 2nd on raw intelligence but last on writing vibes, 5-model comparison shows — iruletheworldmo · 2026-11-21
- Inkling 2 Mini hits 613 tok/s decode and 1M context on one RTX 5090 via Basalt — jesdga95 · 2026-11-21
- Single-Shot Benchmarks Are Dead: Evals Are Now a Compute and Protocol Game — Gauri_the_great · 2026-11-21
- OpenAI releases prompt templates to fix GPT-6.2 Astra and Sol quirks — The Decoder · 2026-11-21
- Unverified leak: 32B, 122B and 256B open-weight models incoming for local AI — HankYeomans · 2026-11-21
- Dev's take: Gemini beats Claude on UI/UX code, Codex 'the absolute worst' — HankYeomans · 2026-11-21