Claude Opus 4.6

2026-02-05 · Publisher : Anthropic · Domain : Artificial Intelligence

Anthropic launches Claude Opus 4.6, its most capable model to date, better at coding, long agentic tasks and reliability on large codebases, with a 1M-token window in beta.

### Specifications - Context window: 1M tokens in beta at launch (out of beta since March 13, 2026, at standard pricing). - Maximum output: 128,000 tokens. Text and image input, text output. - Knowledge cutoff: not disclosed in the pages consulted. - Availability: claude.ai, API, major clouds. API identifier: `claude-opus-4-6`. ### Pricing | Item | Price (USD per million tokens) | | --- | --- | | Input | 5 | | Output | 25 | | Input above 200k tokens (at announcement) | 10 | | Output above 200k tokens (at announcement) | 37.50 | | Cache write 5 min / 1 h | 6.25 / 10 | | Cache read | 0.50 | | Batch (input / output) | 2.50 / 12.50 | The long-context tier was removed when the 1M-token window left beta (March 13, 2026). ### Benchmarks Scores published by the vendor (absolute figures limited in the announcement text, the rest is in charts). | Evaluation | Score | | --- | --- | | Humanity's Last Exam (with tools) | 53.0% | | BrowseComp (multi-agent harness) | 86.8% | | MRCR v2, 8 needles, 1M | 76% (Sonnet 4.5: 18.5%) | | BigLaw Bench | 90.2% | | GDPval-AA | about +144 Elo versus GPT-5.2, +190 versus Opus 4.5 | Terminal-Bench 2.0: best score in the industry according to the vendor, value not recorded. ### Key takeaways Opus 4.6 introduces adjustable effort, compaction and adaptive thinking, which shape the rest of the lineup. The price stays that of Opus 4.5.

Tags : LLM, Anthropic, Claude, Opus, Agents, Codage