Claude Sonnet 5.5: price, context window and specs

What Anthropic says it's for

Anthropic's announcement calls Sonnet 5.5 a faster, lower-cost complement to Opus 5.5. It says Sonnet 5.5 runs more than 30% faster than Sonnet 5 and costs up to 30% less per task, mainly because it needs fewer tokens. Anthropic reports 70.6% on Terminal-Bench 4.0 (Sonnet 5: 10.3%) and a GDPval-AA score two points below Opus 5.5, while noting that Opus 5.5 remains clearly stronger at complex, open-ended work. These are Anthropic's numbers.

Pricing details

• Base: $2 per million input tokens and $10 per million output tokens.
• Prompt caching: 5-minute writes $2.50, 1-hour writes $4, cache reads $0.10 (5% of the input price).
• On 7 October 2026 Anthropic halved Sonnet 5.5 cache reads from $0.20 to $0.10. ClaudeDevs said this makes Sonnet 5.5 about 20% cheaper on most agentic work, and that it's API only, with no change to Claude Code usage limits.
• Batch API: 50% off ($1 / $5). The full 1M context is billed at the standard rate.

Changes that can break existing API code

Anthropic lists five breaking changes for code written for Sonnet 5. To turn off up-front thinking you now send thinking type between_tools instead of disabled, and only at high effort or below. Forced tool use returns an error, thinking blocks are tied to the model and conversation, the older computer_20251124 tool isn't accepted on the Claude API or Google Cloud, and the advisor tool rejects some older models as advisors. Setting temperature, top_p or top_k to a non-default value also returns a 400 error. Details are in the model page and release notes.

Using Sonnet 5.5 in Claude Code

• Claude Code v2.1.284 (28 Sep 2026) made claude-sonnet-5-5 the default Sonnet model on the Anthropic API.
• The sonnet alias resolves to Sonnet 5.5 only on the Anthropic API. On Claude Platform on AWS it resolves to Sonnet 4.6, and on Bedrock, Google Cloud and Microsoft Foundry to Sonnet 4.5, per the model configuration docs. Pick the full model name there to get 5.5.
• Switch with /model sonnet. Claude Code starts Sonnet 5.5 at medium effort.
• Anthropic's Claude Code usage guide suggests Sonnet for most coding work, and planning with Opus then executing with Sonnet, which is what /model opusplan does.

Independent benchmarks

Artificial Analysis lists Claude Sonnet 5.5 (max effort, default fallback) at 56 on its Intelligence Index v4.3.2, at $5.46 per Intelligence Index task, as of 13 October 2026. In its 12 October post, Claude Code with Sonnet 5.5 scored 68 on its Coding Agents Index, level with Meta's Muse Wonder Full in Muse Code and ahead of Claude Code with Opus 5.5 at 66.