AI Research Atlas

Claude Sonnet 5

Anthropic · 30 June 2026

Claude Sonnet 5 is the most agentic Sonnet, near Opus 4.8 on agentic evals, default for Free and Pro, at $2/$10 per million tokens.

Large gains over Sonnet 4.6 in reasoning, tool use, coding and knowledge work; at higher effort it matches Opus 4.8 on some tasks. 1M context, adaptive thinking by default. Lower rate of undesirable behaviours than Sonnet 4.6 and far weaker cyber ability than the Opus models. Introductory $2/$10 pricing (list $3/$15) was made permanent 2026-08-10.

Date
Tuesday, 30 June 2026
Lab
Anthropic
Kind
model
Access
closed API
Price
$2/$10 per M tokens in/out (permanent from 2026-08-10)

Figures

MeasureValueMeasured by
Price per million tokens$2 in / $10 out
introductory price made permanent 2026-08-10; list price had been $3/$15
company

Launch charts (BrowseComp, OSWorld-Verified) are images; a BrowseComp chart was corrected on 2026-06-30 after a methodology error. Sonnet 5.5's launch table later restates Sonnet 5 at 10.3% on Terminal-Bench 4.0 and 57.0% on OSWorld 2.1 (Anthropic runs). Default model on Free and Pro at launch.

Sources

  1. www.anthropic.com/news/claude-sonnet-5
  2. platform.claude.com/docs/en/release-notes/overview

This record was checked against its sources on 6 October 2026. How we check

Related