Anthropic
Sonnet 4.6
- Input / output per 1M tokens
- $3 / $15
- Context (beta)
- 1M
In focusRelease
Anthropic releases Claude Sonnet 4.6 (opens in a new tab)
The price-performance point most agent workloads will actually run on.
anthropic.com ·
Opus 4.6 and GPT-5.3-Codex ship the same day, Sonnet 4.6 becomes the new workhorse, Qwen3.5 adds open weights, and the Pentagon labels Anthropic a supply chain risk.
Anthropic
Sonnet 4.6
The price-performance point most agent workloads will actually run on.
A premium tier kicks in above 200k, which matters if you feed it whole repos.
Steerable mid-task instead of fire-and-forget, which changes how you supervise a long run.
The month's biggest open-weights drop for self-hosters.
Available in the Gemini API, Vertex AI, and the CLI at once, and in Copilot preview the same day.
An agent that runs what it builds hands you evidence instead of a diff to trust.
Execution and the filesystem stay on your machine, so the session moves without the code doing so.
If you sell into government-adjacent supply chains, model choice is now a compliance question.
The practical fallout is tightened API access controls and stricter account verification.
The concrete part for builders is infrastructure, not the declaration.
Directly relevant for agents that should accumulate reusable skills instead of relearning every time.