Claude Opus 4.6 Released: Anthropic's Most Powerful Model Yet
Anthropic has released Claude Opus 4.6, and it is a meaningful upgrade over Opus 4.5. The new model delivers improved reasoning depth, better tool use, and enhanced coding capabilities that cement Claude's position as the leading model for complex agentic workflows.
What is New in Opus 4.6
- Enhanced reasoning: Deeper chain-of-thought capabilities with fewer logical errors on complex multi-step problems
- Improved coding: Better understanding of large codebases, more reliable multi-file edits, and improved test-writing ability
- Better tool use: More reliable function calling, shell command execution, and file operations -- critical for Claude Code workflows
- Extended thinking improvements: The thinking mode produces more thorough and accurate reasoning chains
- Reduced hallucination: Measurably fewer factual errors and fabricated details in outputs
Benchmark Performance
| Benchmark | Opus 4.6 | Opus 4.5 | GPT-5.2 |
|---|---|---|---|
| SWE-bench Verified | 82.1% | 77.4% | 78.3% |
| MMLU | 93.2% | 92.1% | 91.8% |
| HumanEval+ | 93.8% | 91.3% | 90.7% |
| MATH | 90.1% | 88.7% | 87.9% |
Claude Code Gets Even Better
The biggest practical impact of Opus 4.6 is on Claude Code. Since Claude Code uses Opus as its primary model for complex tasks, the improvements translate directly to better agentic coding:
- More reliable multi-file refactors that maintain consistency across the codebase
- Better understanding of project structure and conventions
- Fewer instances of the agent getting "stuck" in loops or making circular edits
- Improved git commit messages and PR descriptions
Pricing
Anthropic has kept pricing identical to Opus 4.5:
- Input: $15.00 per million tokens
- Output: $75.00 per million tokens
- Context window: 200K tokens
- Prompt caching: 50% discount on cached input tokens
This is a smart move by Anthropic -- a capability upgrade at the same price is the best kind of update for existing customers.
Who Should Upgrade
If you are currently using Opus 4.5, switching to 4.6 is a no-brainer. Same price, better results. For teams deciding between Opus 4.6 and GPT-5.2, the choice depends on your workflow:
- Choose Opus 4.6 if: You use Claude Code, need agentic capabilities, or prioritize coding quality
- Choose GPT-5.2 if: You need the broadest general intelligence or are deep in the OpenAI ecosystem
Use our Token Calculator to estimate costs for your specific usage patterns, and see all model options on our models page.