Claude Opus 4.6 in Claude Code: tokens per question
Opus is Claude's higher-priced general-purpose tier, often chosen when a coding task needs sustained reasoning.
The figures
- Bare Claude Code
- 45,700
- With capsul
- 16,400
- less input sent
- x2.78
- answer checks passed out of five, bare then with capsul
- 4 / 4
- tool steps per question, bare then with capsul
- 4.4 / 3.4
- output tokens per question, bare then with capsul
- 1,420 / 425
- API-equivalent cost per question, bare then with capsul
- $0.332 / $0.101
cache-weighted input tokens per question
cache-weighted input tokens per question
90% interval: x1.70 to x4.60
This earlier Opus release has a higher API list price than Sonnet and a lower one than Fable. That makes repeated exploration particularly worth watching: each extra tool step adds context that can be sent again later in the session. Reserve this tier for the investigations, design decisions and difficult edits where its capability is useful.
The table measures short questions, not a full development day. Its question-level rows show whether the result depends on a particular task, while the answer checks show whether the agent found the required evidence. Read both before turning the displayed gain into a budget assumption.
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 | 38,400 | 20,600 | x1.87 | 0 / 0 |
| Q2 | 32,700 | 21,600 | x1.51 | 2 / 3 |
| Q3 | 30,700 | 9,530 | x3.23 | 3 / 3 |
| Q4 | 90,300 | 18,300 | x4.94 | 3 / 3 |
| Q5 | 36,500 | 12,200 | x2.99 | 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 4.6 use per question in Claude Code?
In the September 22, 2026 benchmark, a bare Claude Code session on Claude Opus 4.6 sent 45,700 cache-weighted input tokens per one-line coding question on average. With capsul on the same model, 16,400.
How much does a coding question cost on Claude Opus 4.6?
At Anthropic list prices, the bare agent's traffic came to $0.332 per question and capsul's to $0.101. 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 4.6 established?
Yes. The 90% interval runs from x1.70 to x4.60, entirely above one, on 30 measured cells.
Does capsul change Claude Opus 4.6's answers?
Not on this protocol: Claude Opus 4.6 passed 4 of five answer checks bare and 4 of five with capsul.