Claude Opus 4.6
Release
Anthropic released Claude Opus 4.6 on February 5, 2026. It is the successor to Claude Opus 4.5 (November 24, 2025) and the latest in the Opus line of frontier models.
Pricing remains unchanged at $5/$25 per million input/output tokens. The model is available on claude.ai, the Claude API (claude-opus-4-6), and major cloud platforms.
Key Features
1M Token Context Window (Beta) First Opus-class model to offer a 1 million token context window. On the 8-needle 1M variant of MRCR v2, Opus 4.6 scores 76% versus Sonnet 4.5’s 18.5%.
Agent Teams (Claude Code) Claude Code now supports assembling agent teams â multiple agents working on tasks together.
Compaction API-level context summarization. The model can compress its own context to sustain longer-running tasks without hitting token limits.
Adaptive Thinking Replaces the binary extended thinking toggle. The model adjusts reasoning depth based on contextual cues. Developers can also set explicit effort levels (low/medium/high) via the API.
Effort Controls
New /effort parameter for fine-grained control over intelligence, speed, and cost tradeoffs. Default is high.
Claude in Excel + PowerPoint Upgraded Excel integration. PowerPoint support released in research preview.
Benchmarks
| Benchmark | Result | Notes |
|---|---|---|
| Terminal-Bench 2.0 | #1 | Agentic coding evaluation |
| Humanity’s Last Exam | #1 | Multidisciplinary reasoning |
| GDPval-AA | +144 Elo vs GPT-5.2 | Finance, legal, knowledge work |
| BrowseComp | #1 | Hard-to-find information retrieval |
| BigLaw Bench | 90.2% | Legal reasoning (highest Claude score) |
| MRCR v2 (8-needle, 1M) | 76% | Long-context retrieval |
Opus 4.6 outperforms its predecessor (Opus 4.5) by 190 Elo points on GDPval-AA.
Safety
Anthropic reports Opus 4.6 maintains alignment levels equal to or better than Opus 4.5 across automated behavioral audits. The model shows the lowest over-refusal rate of any recent Claude model.
New safety measures include six cybersecurity probes to detect potential misuse of the model’s enhanced security capabilities. Anthropic also published a system card covering all capability and safety evaluations.
Context
This is the 12th major Claude release since the original Claude 1.0 in March 2023. The release continues a pattern of roughly quarterly Opus-class updates throughout 2025â2026:
- Claude Opus 4 â May 22, 2025
- Claude Opus 4.1 â August 5, 2025
- Claude Opus 4.5 â November 24, 2025
- Claude Opus 4.6 â February 5, 2026
Timeline Update
This release has been added to the whatis ai timeline, along with 13 other model releases from 2025 that were previously missing from the tracker.
References
Configuration details reflect a production environment at time of writing. Implementation specifics vary based on tooling versions, platform updates, and organizational requirements. Validate approaches against current documentation before deployment.