Claude Opus 4
2025-05-22 · Publisher : Anthropic · Domain : Artificial Intelligence
Anthropic launches Claude Opus 4, presented as its best coding model, capable of long-running tasks spanning thousands of steps. It ships alongside Claude Sonnet 4 and the general availability of Claude Code.
What changes
- Positioned as the top-performing model in the lineup for coding (72.5% on SWE-bench Verified, 43.2% on Terminal-bench).
- Sustained work on long tasks: Rakuten cites a 7-hour autonomous run.
- Extended thinking with tool use (beta), parallel tool execution, better instruction following.
- Memory capabilities: with local file access, the model creates and maintains memory files.
- New API features at launch: code execution tool, MCP connector, Files API, prompt caching for up to one hour.
Anthropic launches Claude Opus 4, presented as its best coding model, capable of long-running tasks spanning thousands of steps. It ships alongside Claude Sonnet 4 and the general availability of Claude Code. ### What changes - Positioned as the top-performing model in the lineup for coding (72.5% on SWE-bench Verified, 43.2% on Terminal-bench). - Sustained work on long tasks: Rakuten cites a 7-hour autonomous run. - Extended thinking with tool use (beta), parallel tool execution, better instruction following. - Memory capabilities: with local file access, the model creates and maintains memory files. - New API features at launch: code execution tool, MCP connector, Files API, prompt caching for up to one hour. ### Specifications - API identifier: claude-opus-4-20250514 (retired from the API on June 15, 2026). - Availability: Claude.ai (Pro, Max, Team, Enterprise), Anthropic API, Amazon Bedrock, Google Cloud Vertex AI. - Context window, maximum output and knowledge cutoff: not disclosed in the announcement (200K context and 32K output reported in secondary sources). - Modalities: text and images as input, text as output. ### Pricing | Input (USD / M tokens) | Output (USD / M tokens) | 5-min cache write | Cache read | Batch | |---|---|---|---|---| | 15 | 75 | 18.75 | 1.50 | 7.50 / 37.50 | Cache and batch prices come from the current pricing page of the documentation. ### Benchmarks | Benchmark | Claude Opus 4 | Claude Sonnet 4 | |---|---|---| | SWE-bench Verified | 72.5% | 72.7% | | SWE-bench (high compute) | 79.4% | 80.2% | | Terminal-bench | 43.2% | not included | | GPQA Diamond (extended thinking) | 75.8% | 71.4% | | MMMLU (extended thinking) | 88.0% | 85.8% | | MMMU (extended thinking) | 74.6% | 73.3% | | AIME (extended thinking) | 34.5% | 33.5% | Scores published by the vendor. ### Key takeaways Opus 4 targets long-running coding and agent tasks, at a high price (15 / 75 USD). It will be replaced at this price point by Opus 4.1, then by Opus 4.5 at a third of the price.
Links
Tags : LLM, Anthropic, Claude, Opus, Codage, Agents