Last updated:
Claude Sonnet 5.5: pricing, cost per task and benchmarks
Anthropic released Claude Sonnet 5.5 on September 28, 2026. The list price did not change, so the real question is what a coding task costs now and when Opus 5.5 is still worth double.
Key facts
| API model ID | claude-sonnet-5-5 |
|---|---|
| Released | September 28, 2026 |
| Context window | 1M tokens |
| Max output | 128K tokens |
| Knowledge cutoff | June 2026 |
| Available on | Claude API, AWS, Google Cloud, Microsoft Azure, Claude apps and Claude Code |
| Default effort | Medium in Claude Code and the apps, High on the API |
Source: Anthropic: Claude models overview
Claude Sonnet 5.5 API pricing
List prices in USD per 1M tokens, from Anthropic's pricing page. Sonnet 5 has exactly the same prices.
| Model | Input | Output | Cache write (5 min) | Cache hit |
|---|---|---|---|---|
| Claude Sonnet 5.5 | $2.00 | $10.00 | $2.50 | $0.20 |
| Claude Opus 5.5 | $4.00 | $20.00 | $5.00 | $0.20 |
| Claude Haiku 4.5 | $1.00 | $5.00 | $1.25 | $0.10 |
The Batch API takes 50% off input and output for jobs that can wait.
Source: Anthropic: Claude API pricing
What one coding task really costs
Using the same assumptions as our calculator (60% of input served from cache), here is one agent task by codebase size. Sonnet 5.5 often needs fewer tokens than this, so treat it as an upper estimate.
| Codebase size | Sonnet 5.5 | Opus 5.5 | Muse Code (Standard) |
|---|---|---|---|
| Small (< 10k lines) | $0.115 | $0.223 | $0.061 |
| Medium (10k-100k lines) | $0.258 | $0.498 | $0.140 |
| Large / monorepo (> 100k lines) | $0.618 | $1.188 | $0.342 |
At 173 tasks a month (8 a day, 5 days a week, medium codebase) that is about $45 a month on Sonnet 5.5 versus $86 on Opus 5.5, before the fewer-tokens gain.
Compare it with Claude Code plans and other tools in the cost calculator
Which effort level to use
- Anthropic sets Medium as the default in Claude Code and the apps, and High on the API.
- ComputingForGeeks ran the same 9 generation calls at each level: Medium cost $0.0665, High $0.1230 and Max $1.3585, about 20x Medium, with identical pass results.
- On Anthropic's FrontierCode 1.1 numbers, Max scored 46.2%, below xhigh at 52.1%. Vellum attributes this to extra subagents hitting timeouts.
- Start on Medium, move to High for hard bugs, and keep Max for rare cases you have tested.
Source: Anthropic: Introducing Claude Sonnet 5.5
Benchmarks (Anthropic's numbers)
From Anthropic's launch post. Opus 5.5 is shown at its highest effort setting.
| Benchmark | Sonnet 5.5 | Sonnet 5 | Opus 5.5 |
|---|---|---|---|
| Terminal-Bench 4.0 | 70.6% | 10.3% | 66.4% |
| FrontierCode 1.1 (xhigh) | 52.1% | 42.4% | 54.4% |
| CursorBench 4.0 | 55.5% | 34.1% | 57.8% |
| OSWorld 2.1 | 80.1% | 57.0% | 81.8% |
| Humanity's Last Exam | 64.5% | 54.9% | 67.7% |
| GDPval-AA v2.1 (Elo) | 1844 | 1449 | 1846 |
- Anthropic notes some third-party scores (GDPval-AA, AA-Briefcase) were run on a pre-release deployment with a bug affecting structured outputs.
- Anthropic also says Opus 5.5 remains clearly stronger at difficult, open-ended work, and that benchmarks are not direct proxies for every production workload.
Source: Anthropic: Introducing Claude Sonnet 5.5
What early reviews found
- ComputingForGeeks: Hands-on test in Claude Code: fixing the same bugs took Sonnet 5.5 3 tool calls, $0.057 and 12 seconds per run, against 12-13 calls, $0.157 and 48 seconds for Sonnet 5.
- VentureBeat: Collects launch-partner reports: Box saw 2.4x faster runs with 12% fewer tokens, Slack about 14% fewer output tokens, and Lovable about one-third fewer tool calls.
- The Decoder: Points out that the speed and cost-per-task claims are Anthropic's and still need independent confirmation.
- The Next Web: Covers the new cyber safeguards: higher-risk security requests visibly fall back to Sonnet 5, which matters if you do security work.
- Vellum: Calls it a "high-speed workhorse" and flags the Max-below-xhigh result on FrontierCode.
Should you switch?
- Using Sonnet 5 on the API: yes. The price is the same and the model ID is the only change.
- Using Opus 5.5 for routine coding: try Sonnet 5.5 first. It costs half as much per token and is close on most coding benchmarks.
- On a Claude Pro or Max plan: you pay a flat price, so the gain shows up as more work per usage window rather than a lower bill.
- Doing penetration testing or exploit research: expect some requests to fall back to Sonnet 5.
vs Claude Code · Cost calculator
FAQ
How much does Claude Sonnet 5.5 cost?
$2.00 per 1M input tokens and $10.00 per 1M output tokens on the Claude API. Cache hits cost $0.20 per 1M. That is the same as Sonnet 5 and half the price of Opus 5.5 ($4.00 / $20.00).
Is Claude Sonnet 5.5 cheaper than Sonnet 5?
Not per token: the list price is identical. Anthropic says it costs up to 30% less per task because it uses fewer tokens and tool calls. One hands-on test measured $0.057 per run versus $0.157 for Sonnet 5.
What is the Sonnet 5.5 model ID?
claude-sonnet-5-5. Context window 1M tokens, max output 128K tokens.
Can I use Sonnet 5.5 in Claude Code?
Yes. Anthropic lists Claude Code among the launch platforms, with Medium as the default effort there. You can pay through a Claude plan or per token with an API key.
Sonnet 5.5 or Opus 5.5 for coding?
For most day-to-day agent work, Sonnet 5.5: it scored 70.6% on Terminal-Bench 4.0 against 66.4% for Opus 5.5 at half the price. Anthropic still recommends Opus 5.5 for the hardest open-ended problems.
Independent page. Not affiliated with Anthropic. Benchmark figures are Anthropic's or the named reviewer's; we have not run our own tests.