Anthropic has released Claude Opus 4.6, a significant upgrade to their most capable AI model, featuring a 1 million token context window and substantially improved coding abilities.
What's New
Claude Opus 4.6 pushes the boundaries of what large language models can handle in a single session. The headline features include:
- 1M token context window (beta) — available on the Claude Developer Platform
- 128k output tokens — enabling much longer generated responses
- Improved agentic task performance — more reliable on sustained, multi-step workflows
- Enhanced code review and debugging — better at working with large codebases
Benchmark Performance
The model sets new state-of-the-art results across several key benchmarks:
- Terminal-Bench 2.0 — highest score for agentic coding tasks
- Humanity's Last Exam — leading performance on multidisciplinary reasoning
- GDPval-AA — outperforms GPT-5.2 by approximately 144 Elo points
- MRCR v2 — 76% on long-context retrieval, compared to 18.5% for Sonnet 4.5
Developer Features
Anthropic has introduced several new capabilities aimed at developers building on the platform:
- Adaptive thinking — the model autonomously determines when extended reasoning is beneficial, removing the need for manual prompt tuning
- Effort controls — four levels (low, medium, high, max) let developers balance intelligence, speed, and cost
- Context compaction — older context is automatically summarised during long tasks, keeping the model effective over extended sessions
- Agent teams — multiple agents can now work collaboratively within Claude Code
Safety
Anthropic reports that Opus 4.6 shows low rates of misaligned behaviour and the lowest over-refusal rates among recent Claude versions, meaning it is less likely to refuse legitimate requests while still maintaining safety guardrails.
Pricing and Availability
Claude Opus 4.6 is available now on claude.ai, the Anthropic API (model ID: claude-opus-4-6), and major cloud platforms. Standard pricing remains at $5 / $25 per million input / output tokens, with a premium tier of $10 / $37.50 for inputs exceeding 200k tokens.