Claude Sonnet 4.5 sustains 30-hour tasks

Monday 29 September 2025 · significance ●●●○○

Anthropic reports Claude Sonnet 4.5 maintaining focus for more than 30 hours on complex multi-step tasks, and 61% on OSWorld.

What happened and why it matters

Sonnet 4.5 led SWE-bench Verified (77.2%) and jumped from 42% to 61% on OSWorld in four months. Anthropic released the Claude Agent SDK, the infrastructure behind Claude Code, alongside it.

Sources

Introducing Claude Sonnet 4.5 Anthropic · 29 Sep 2025

Effect on the Gapnium Coefficient

+0.0052

0.2576 → 0.2628

Adjustments

gapnium.autonomy +0.020 Day-plus unsupervised runs; large OSWorld jump.
gapnium.recursion +0.006 Agent infrastructure used internally to build AI is opened to everyone.

Review

Last reviewed by AI agent: 7 Nov 2026
Model: seed:claude-opus-5-5
Agent confidence: 0.75
Human reviewed: —

Initial historical seed. Scores assigned by the methodology rubric; awaiting first scheduled re-review.

Feedback on this scoring

Hover the chart to read values · click a marker to open its articles