CONFIRMEDDate confidence: CONFIRMEDStated as firmHow firmly the source words the date: confirmed, scheduled, estimated, delayed or unverified.Show all CONFIRMED pins89% CONFIDENCEOverall confidence: 89%How well the pin's source and references back up its dates.Weighted average of how firmly 6 references, the source included, support the pin's start and end times; a reference counts half as much for every 180 days older than the newestShow all pins at 75% confidence or better
Anthropic's[1][3] post is dated "Aug 5, 2025": "Today we're releasing Claude[2][4] Opus 4.1 ... now available to paid Claude users and in Claude Code. It's also on our API, Amazon Bedrock, and Google Cloud's Vertex AI"; the Claude release notes and GitHub's[5] changelog carry the same date.- Start
- Aug 5, 202590% CONFIDENCEsource
Claude Opus 4.1 Released
Anthropic releases Claude Opus 4.1, an upgrade to Claude Opus 4 that raises its SWE-bench Verified coding score to 74.5% at the same price, on the day OpenAI released its first open-weight models since 2019.
- Anthropic released Claude Opus 4.1 on 5 August 2025 to paid Claude users, in Claude Code, and on its API, Amazon Bedrock and Google Cloud's Vertex AI, at the same price as Opus 4[1][2]
- It came out the same day OpenAI released its first open-weight reasoning models since 2019; Anthropic said it planned "substantially larger improvements to our models in the coming weeks"[1][6]
- GitHub put it into Copilot the same day as a public preview for Enterprise and Pro+ plans[5]
- Like Opus 4 it is deployed under the ASL-3 safety standard[3]
- Retired from the Claude API on 5 August 2026, exactly a year after launch, replaced by Claude Opus 4.8[4]
Notable features
- An upgrade to Claude Opus 4 "on agentic tasks, real-world coding, and reasoning", with better in-depth research and data analysis, "especially around detail tracking and agentic search"[1]
- 74.5% on SWE-bench Verified against Opus 4's 72.5%, and 43.3% on Terminal-Bench against 39.2%[1][6]
- Other scores in Anthropic's table: 80.9% on GPQA Diamond, 82.4% on TAU-bench retail, 89.5% on MMMLU and 78.0% on AIME 2025[1]
- Customer reports: GitHub sees notable gains in multi-file refactoring, Rakuten precise fixes in large codebases "without making unnecessary adjustments", and Windsurf a one-standard-deviation jump over Opus 4[1]
- Priced like Opus 4 at $15 per million input and $75 per million output tokens; API id claude-opus-4-1-20250805[1][6]
Benchmarks[1]
| Benchmark | Claude Opus 4.1 | Claude Opus 4 | Claude Sonnet 4 | OpenAI o3 | Gemini 2.5 Pro |
|---|---|---|---|---|---|
| SWE-bench Verified¹ (agentic coding) | 74.5% | 72.5% | 72.7% | 69.1% | 67.2% |
| Terminal-Bench² (agentic terminal coding) | 43.3% | 39.2% | 35.5% | 30.2% | 25.3% |
| GPQA Diamond (graduate-level reasoning) | 80.9% | 79.6% | 75.4% | 83.3% | 86.4% |
| TAU-bench (agentic tool use) | Retail 82.4%, Airline 56.0% | Retail 81.4%, Airline 59.6% | Retail 80.5%, Airline 60.0% | Retail 70.4%, Airline 52.0% | — |
| MMMLU³ (multilingual Q&A) | 89.5% | 88.8% | 86.5% | 88.8% | — |
| MMMU, validation (visual reasoning) | 77.1% | 76.5% | 74.4% | 82.9% | 82% |
| AIME 2025⁴ (high school math competition) | 78.0% | 75.5% | 70.5% | 88.9% | 88% |
¹ Opus 4.1, Opus 4 and Sonnet 4 run pass@1 with bash/editor tools, averaged over 10 trials, single-attempt patches, no test-time compute. ² Default agent framework (Terminus 1), averaged over 5 trials. ³ Claude scores are the average over 14 non-English languages. ⁴ Run with nucleus sampling, top_p 0.95.
Buy onClaude