Claude
Worth testingConsider it for iterative document revision with explicit constraints and reusable project context.
Compare the announced Haiku and Opus API rates, context-dependent Haiku pricing and suggested evaluation tasks without treating benchmarks as your result.
Haiku 5.5 and Opus 5.5 are different members of Anthropic's Claude family. The Haiku announcement emphasizes narrow, high-volume work; the Opus announcement emphasizes complex coding and knowledge work. Choose a task and billing route before comparing headline prices.
USD per million tokens, checked October 30, 2026:
| Model / prompt length | Input | Output | Cache read |
|---|---|---|---|
| Haiku 5.5, up to 100,000 prompt tokens | $0.10 | $0.50 | $0.01 |
| Haiku 5.5, over 100,000 prompt tokens | $0.50 | $2.50 | $0.05 |
| Opus 5.5 | $4.00 | $20.00 | $0.20 |
The announcements also describe cache-write and other billing details. The table is not an all-in project quote: request length, caching, generated tokens, retries and provider terms affect the bill. Claude app-plan allowances and direct API charges are separate access questions.
| Trial task | What to measure |
|---|---|
| Classify a small batch of known examples | Correct labels and costly mistakes |
| Summarize a document | Coverage and support for each important claim |
| Make a repository change | Behavior, tests and necessary corrections |
Use the same inputs and acceptance criteria, retain the exact model and settings, and count the total cost to reach an acceptable result. A lower per-token rate does not guarantee a lower cost per finished task; a larger model's benchmark score does not establish a benefit on every small task.
The launch posts report provider benchmarks and customer experiences. Varwo has not reproduced those evaluations or run paid requests. For product-plan context, see Claude; for an external coding subscription, compare the separate OpenCode Go route.
Check that a tool accepts the inputs you need before you pay.
Consider it for iterative document revision with explicit constraints and reusable project context.