Claude Sonnet 5.5 vs Claude Opus 5.5 cost.
Anthropic shipped both within a week. Sonnet 5.5 is half the price per token and, on the main independent measurement, more expensive per finished task.
- Input gap
- 2×
- Output gap
- 2×
- Cached gap
- same
- Rates verified
- 2026-10-02
Current-model note: Both models are current. Claude Opus 5.5 was released on 22 September 2026 and Claude Sonnet 5.5 on 28 September 2026. Rates come from the maintained catalogue.
| Current model | Input / 1M | Cached / 1M | Cache write / 1M | Output / 1M | Context |
|---|---|---|---|---|---|
| Claude Sonnet 5.5Official pricing ↗ | $2.00 | $0.20 | $2.50 | $10.00 | 1,000,000 |
| Claude Opus 5.5Official pricing ↗ | $4.00 | $0.20 | $5.00 | $20.00 | 1,000,000 |
Rates verified 2026-10-02. The browser may refresh them from the live catalogue. Provider pricing and tiers remain authoritative.
Claude Sonnet 5.5: Provider-Calibrated UTF-8 Projection
Claude Opus 5.5: Provider-Calibrated UTF-8 Projection
PDF, DOCX, code, text, and data files are extracted locally and applied to both models.
Claude Sonnet 5.5: documented formula
Claude Opus 5.5: documented formula
Prompt text, document contents, and image pixels stay in the browser.
Text, code, PDF, DOCX, images, audio, or video · 12 MB per file
Your input, measured across providers.
Choose a provider card to inspect every available model.
Workload & Scaling Planner +
What five real workloads actually cost
Headline per-token prices rarely decide a bill. Request shape does. These figures are calculated from the verified rates for Claude Sonnet 5.5 and Claude Opus 5.5, including any long-context tier that applies once a request crosses its threshold.
| Workload | Input | Output | Claude Sonnet 5.5 | Claude Opus 5.5 | Difference |
|---|---|---|---|---|---|
| Short chat turn A typical assistant exchange. | 1,000 | 500 | $0.0070 | $0.014 | 2.00x cheaper on Claude Sonnet 5.5 |
| RAG answer Five retrieved chunks plus a question. | 12,000 | 800 | $0.032 | $0.064 | 2.00x cheaper on Claude Sonnet 5.5 |
| Code review A medium pull request with surrounding files. | 60,000 | 2,000 | $0.140 | $0.280 | 2.00x cheaper on Claude Sonnet 5.5 |
| Whole-document analysis A long report or contract read in one call. | 300,000 | 4,000 | $0.640 | $1.28 | 2.00x cheaper on Claude Sonnet 5.5 |
| Full-context load Filling most of a one-million-token window. | 900,000 | 4,000 | $1.84 | $3.68 | 2.00x cheaper on Claude Sonnet 5.5 |
"(tier)" marks a request that crossed a long-context threshold, so it is billed above the headline rate. Output length is the assumption most worth changing for your own case: paste a real prompt into the calculator above to replace these with your numbers.
The cached-input rate is the number most comparisons miss
Agents, chat threads and RAG pipelines resend the same system prompt and context on every call. Those repeated tokens bill at the cached rate, not the headline rate, so a model can be cheaper on paper and more expensive in production.
| Model | Fresh input / 1M | Cached input / 1M | Discount | Break-even |
|---|---|---|---|---|
| Claude Sonnet 5.5 | $2.00 | $0.200 | 90% | 1 read |
| Claude Opus 5.5 | $4.00 | $0.200 | 95% | 1 read |
Break-even assumes a cache write costs about 1.25x the base input rate, which is what OpenAI and Anthropic currently document for a short time-to-live. It answers one question: how many times a cached prefix must be re-read before caching is cheaper than paying full price each call. Above that count, every further read saves the discount shown.
Half the price per token
Claude Sonnet 5.5 lists $2 per million input tokens, $0.20 cached and $10 output. Claude Opus 5.5 lists $4, $0.20 and $20. On the sticker, Sonnet 5.5 is half the price on input and output, and the two are identical on cached reads.
Opus 5.5 is also cheaper than the Opus 5 it replaces, which listed $5 and $25, and its cache reads fell from a tenth of the input price to a twentieth. The Opus tier became about 20 percent cheaper overnight.
Cache writes cost $2.50 per million on Sonnet 5.5 and $5 on Opus 5.5 for a five-minute lifetime, and twice that for an hour.
And more expensive per finished task
Artificial Analysis measures the cost of running its whole Intelligence Index, which counts every token a model spends to reach its answers. On version 4.3.2 it scored Claude Opus 5.5 at 58, the highest of the 224 models it ranks, and Claude Sonnet 5.5 at 56, second.
The cost of those runs went the other way: $5.98 for Opus 5.5 and $7.67 for Sonnet 5.5. The reason is output volume. Sonnet 5.5 at maximum effort emitted about 193,000 output tokens per task, the most Artificial Analysis has recorded. A model at half the rate that writes three times as much is not cheaper.
Lower the effort setting and the picture changes again: Opus 5.5 at its xhigh setting matched Sonnet 5.5 at maximum for $3.46 against $7.60. Artificial Analysis also notes it tested Sonnet 5.5 on a pre-release build with a structured-outputs bug, so treat the exact figures as a snapshot rather than a verdict.
What to measure instead
Per-token price answers "what does a token cost". It does not answer "what does the job cost", which is the question a budget asks. The bridge between them is the number of output tokens your tasks really need.
Run the same prompt on both, with the effort setting you would ship, and compare total cost per completed task. The calculator below prices the input exactly; set the output length to what you measure in practice rather than to a round number.
Both models count tokens with the newer Claude tokenizer, which produces about 30 percent more tokens for the same text than Claude Sonnet 4.6 and earlier. Counts measured on an older Claude model do not transfer.
Model the workload, not the marketing price.
Do not choose on the per-token price. Sonnet 5.5 writes far more output than Opus 5.5 for the same work, so the model that looks half the price can produce the larger invoice. Measure output length on your own tasks before switching.
Review calculation methodology →Measure against the model you use
Other comparisons worth running
Compare approaches, not just models
Frequently asked questions
Is Claude Sonnet 5.5 cheaper than Claude Opus 5.5?+
Per token, yes: $2 and $10 per million against $4 and $20. Per completed task, Artificial Analysis measured the opposite on version 4.3.2 of its index: $7.67 for Sonnet 5.5 against $5.98 for Opus 5.5, because Sonnet 5.5 writes far more output.
Why does the cheaper model cost more per task?+
Because cost is the token price multiplied by the tokens used. Sonnet 5.5 at maximum effort emitted about 193,000 output tokens per index task, the highest Artificial Analysis has recorded, which more than cancels its lower rate.
How much cheaper is Opus 5.5 than Opus 5?+
About 20 percent per token: $4 and $20 per million against $5 and $25. Cache reads fell further, from a tenth of the input price to a twentieth, which is $0.20 per million.
Do these models count tokens the same way as older Claude models?+
No. Claude 4.7 and later use a newer tokenizer that produces roughly 30 percent more tokens for the same text. Anthropic advises counting your prompt against the model you plan to use rather than reusing an older count.