Claude Opus 4.7 in Claude Code: tokens per question
This Opus release sits in the higher-priced Claude tier between Sonnet and Fable.
The figures
- Bare Claude Code
- 38,700
- With capsul
- 20,100
- less input sent
- x1.92
- answer checks passed out of five, bare then with capsul
- 5 / 4
- tool steps per question, bare then with capsul
- 3.6 / 2.4
- output tokens per question, bare then with capsul
- 856 / 397
- API-equivalent cost per question, bare then with capsul
- $0.28 / $0.126
cache-weighted input tokens per question
cache-weighted input tokens per question
90% interval: x1.26 to x3.14
Its API list price matches the neighbouring Opus releases in this catalogue. The useful comparison is therefore not a nominal price change but how much input each run carried and whether the answer passed its checks. For complex debugging, the cost of finding the right files can exceed the cost of the question itself.
A model can take a different path through the same codebase on another run. The interval and per-question rows expose some of that variation. Use them when deciding whether this Opus version is a sensible default or a model to call only for harder work.
Question by question
Five one-line questions about a real TypeScript codebase, each asked 3 times per arm. Weighted input per question, mean of the repetitions.
| Question | Bare | capsul | Less input | Correct runs, bare / capsul |
|---|---|---|---|---|
| Q1 | 44,200 | 24,500 | x1.80 | 2 / 0 |
| Q2 | 47,700 | 29,700 | x1.61 | 3 / 3 |
| Q3 | 35,500 | 11,900 | x3.00 | 3 / 3 |
| Q4 | 30,500 | 19,200 | x1.59 | 3 / 3 |
| Q5 | 35,400 | 15,300 | x2.31 | 3 / 3 |
List prices
| Model | Input | Output | Cache read | Cache write 5 min | Cache write 1 h | Context |
|---|---|---|---|---|---|---|
| Claude Haiku 4.5 | 1.00 | 5.00 | 0.10 | 1.25 | 2.00 | 200,000 |
| Claude Sonnet 5 | 2.00 | 10.00 | 0.20 | 2.50 | 4.00 | 1,000,000 |
| Claude Sonnet 4.6 | 3.00 | 15.00 | 0.30 | 3.75 | 6.00 | 1,000,000 |
| Claude Opus 4.6 | 5.00 | 25.00 | 0.50 | 6.25 | 10.00 | 1,000,000 |
| Claude Opus 4.7 | 5.00 | 25.00 | 0.50 | 6.25 | 10.00 | 1,000,000 |
| Claude Opus 4.8 | 5.00 | 25.00 | 0.50 | 6.25 | 10.00 | 1,000,000 |
| Claude Opus 5 | 5.00 | 25.00 | 0.50 | 6.25 | 10.00 | 1,000,000 |
| Claude Opus 5.5 | 4.00 | 20.00 | 0.20 | 5.00 | 8.00 | 1,000,000 |
| Claude Fable 5 | 10.00 | 50.00 | 1.00 | 12.50 | 20.00 | 1,000,000 |
| Claude Fable 5.1 | 10.00 | 50.00 | 0.25 | 12.50 | 20.00 | 1,000,000 |
Anthropic list prices, checked September 23, 2026.
How this was measured
Two arms on the same model: Claude Code as anyone runs it, and the same question through capsul. The unit is cache-weighted input: fresh tokens at full weight, cache reads at a tenth. Campaign of September 22, 2026, 3 repetitions per question and arm.
30 / 30 cells · /benchmarks/2026-09-22-claude.json
Questions
How many tokens does Claude Opus 4.7 use per question in Claude Code?
In the September 22, 2026 benchmark, a bare Claude Code session on Claude Opus 4.7 sent 38,700 cache-weighted input tokens per one-line coding question on average. With capsul on the same model, 20,100.
How much does a coding question cost on Claude Opus 4.7?
At Anthropic list prices, the bare agent's traffic came to $0.28 per question and capsul's to $0.126. A Claude subscription does not bill per token: read it as how much of a plan each question uses, compared with other models.
Is the saving on Claude Opus 4.7 established?
Yes. The 90% interval runs from x1.26 to x3.14, entirely above one, on 30 measured cells.
Does capsul change Claude Opus 4.7's answers?
It passed fewer: 5 of five answer checks bare, 4 of five with capsul. The benchmark publishes this rather than leaving it out.