AI Agent StoreANTHROPIC · PROPRIETARY
Claude Opus 4.6
See how this model performed in tests and what it could cost for your tasks.
Thinking setting tested: high · adaptive thinking.
Data reviewed · LiveBench question set: 25 June 2026 · View every source ↗ · Download the data ↓
Add to a custom comparison ↗Results by type of task
Explain these tests ↗| Reasoning | 88.7 / 100 |
|---|---|
| Coding | 78.2 / 100 |
| Agentic coding | 49.0 / 100 |
| Mathematics | 89.3 / 100 |
| Data analysis | 69.9 / 100 |
| Language | 83.3 / 100 |
| Instruction following | 63.3 / 100 |
What the evidence says
Claude Opus 4.6 scores 74.5 on LiveBench, averaging its seven task categories equally. Its highest measured category is mathematics (89.3); its lowest is agentic coding (49.0). Categories differ in difficulty, so these are test results, not universal strengths or weaknesses.
Our General agent profile produces an Agent Fit score of 70.5 by emphasizing reasoning, instruction following, agentic coding, coding and data analysis. See the exact weights or adjust the profile in the interactive comparison.
How people rated its answers
Arena reports 1504 ± 3 from 79,415 votes for claude-opus-4-6-high. Combined with LiveBench, its Evidence Consensus score is 55.6 within our 10 versions and settings tested in both sources. This does not represent a market-wide rank.
API features to verify
The catalog lists tool-calling support and image input. These are declared features, not evaluations of tool accuracy or visual understanding.
This is a proprietary model accessed through a provider. Availability, rate limits, data handling and service tiers depend on the endpoint and account.
Claude Opus 4.6 API pricing
Input is what you send to the model; output is what it produces, including charged thinking tokens. Prices are per million tokens, the small pieces of text the model processes.
Direct/source price observed 2026-10-15; catalog offer rechecked 2026-10-15. Older observations remain dated until verified again.
At the baseline of 1,000 calls, 10,000 input tokens and 2,000 billed output tokens per call, the text API estimate is $100. Reasoning belongs in billed output. Caching, tools, retries, long-context premiums, service tiers, hosting and taxes are excluded.
Trace this model to the source
- LiveBench row ID
claude-opus-4-6-thinking-auto-high-effort- Question-set release
- 2026-06-25 · Source CSV ↗
- Catalog ID
- anthropic/claude-opus-4.6 ↗
- Arena configuration
- claude-opus-4-6-high ↗
- Reviewed snapshot
- 2026-10-15
Compare alternatives
For production decisions, test examples from your own work using the same tools and a clear definition of success. Neither a large context window nor a strong benchmark score guarantees reliability in your own agent.