Claude Sonnet 5.5 Pricing: $2 / $10, the Same as Sonnet 5
Anthropic's new Sonnet keeps Sonnet 5's price, and on its published benchmarks it lands close to Opus 5.5 at half the list price.
Anthropic released Claude Sonnet 5.5 on 28 September 2026 at $2 per million input tokens and $10 per million output tokens, the same as Claude Sonnet 5, which it replaces. Cached input costs $0.20 per million and batch requests are half price. Anthropic says the new model typically needs fewer tokens for the same work, so it costs up to 30% less per task even though the rates have not changed.
Claude Sonnet 5.5 at a glance
| API model ID | claude-sonnet-5-5 |
|---|---|
| Input | $2.00 per 1M tokens |
| Output | $10.00 per 1M tokens |
| Cached input | $0.20 per 1M tokens |
| Cache writes | $2.50 per 1M tokens (5-minute cache), $4 (1-hour cache) |
| Batch | 50% off: $1 / $5 per 1M tokens |
| Context window | 1M tokens |
| Maximum output | 128K tokens |
The live figures, with every other Sonnet model, are on our Claude Sonnet pricing page.
How it compares with the rest of the Claude line-up
| Model | Input / output per 1M | Cached input |
|---|---|---|
| Claude Fable 5.1 | $10.00 / $50.00 | $0.25 |
| Claude Opus 5.5 | $4.00 / $20.00 | $0.20 |
| Claude Sonnet 5.5 | $2.00 / $10.00 | $0.20 |
| Claude Sonnet 5 | $2.00 / $10.00 | $0.20 |
| Claude Haiku 4.5 | $1.00 / $5.00 | $0.10 |
Sonnet 5.5 costs half as much as Opus 5.5 on input and output, but the two charge the same $0.20 for cached input. For agents, where most input tokens are cache reads, the gap between them is smaller than the list prices suggest. Anthropic says a Claude Haiku 5.5 will follow "in the coming weeks"; until then Haiku 4.5 remains the cheapest Claude model.
Performance
These are Anthropic's published results, not our own tests. Anthropic ran Sonnet 5.5 at its highest-scoring effort setting.
| Benchmark | Sonnet 5.5 | Sonnet 5 | Opus 5.5 |
|---|---|---|---|
| Terminal-Bench 4.0 (agentic coding) | 70.6% | 10.3% | 66.4% |
| CursorBench 4.0 (agentic coding) | 55.5% | 34.1% | 57.8% |
| GDPval-AA v2.1 (knowledge work, Elo) | 1844 | 1449 | 1846 |
| OSWorld 2.1 (computer use) | 80.1% | 57.0% | 81.8% |
| Humanity's Last Exam (with tools) | 64.5% | 54.9% | 67.7% |
On most of these, Sonnet 5.5 comes within a few points of Opus 5.5, and it beats it on Terminal-Bench. Anthropic still describes Opus 5.5 as clearly stronger on complex, open-ended work. Benchmark scores are a guide, not a substitute for testing on your own tasks.
On the independent Artificial Analysis Intelligence Index, Sonnet 5.5 scored 56 on 2 October, third of the 224 models listed, behind Opus 5.5 at 58. Artificial Analysis measured it at 139 output tokens per second, against 92 for Opus 5.5. Anthropic says it generates output more than 30% faster than Sonnet 5. It was not yet on the Arena leaderboard.
What it costs in practice
For a coding agent doing 400 tasks a month (about 750M input tokens with 85% served from cache, and 8M output tokens), Sonnet 5.5 costs $433 a month, against $738 on Opus 5.5 and $369 on GPT-6.1 Sol. That assumes the same token counts. If Anthropic's "up to 30% fewer tokens" holds for your work, the same 400 tasks would cost about $303.
For a chat workload without caching, such as 10M input and 2M output tokens a month, it costs $40.00, the same as GPT-6.1 Sol at $40.00 and less than Gemini 3.1 Pro at $44.00.
Against GPT-6.1 Sol and Gemini
GPT-6.1 Sol, released a day later, also costs $2 / $10 per 1M tokens, but charges $0.10 for cached input against Sonnet 5.5's $0.20, so it costs less for cache-heavy agents. For chat without much caching, the two cost the same, and the choice comes down to quality and token efficiency. Gemini 3.8 Flash is cheaper at $0.75 / $3.75 until its promotion ends on 31 December. Compare Sonnet 5.5 with GPT-6.1 Sol or with Gemini 3.8 Flash.
Should you switch?
- From Sonnet 5: yes, for most work. The rates are the same, and Anthropic's benchmarks and early testers report better results with fewer tokens. Sonnet 5 stays available at $2 / $10 while you test.
- From Opus 5.5: try Sonnet 5.5 on well-scoped tasks such as bug fixes, documents and routine agent steps. It costs half as much, though cache-heavy agents save less than half.
- If you run Sonnet with thinking turned off: Sonnet 5.5 rejects the old setting. Switch to the new
between_toolssetting first; Anthropic's migration guide covers the change. - If you work in security: higher-risk cybersecurity requests fall back to Sonnet 5. Routine bug finding and fixing are not affected.
Sources: Anthropic pricing, Anthropic announcement, Artificial Analysis.