Anthropic · released
Claude Opus 5
Top of the intelligence index, at half the cost of the tier above.
Overview
Anthropic's July flagship and, at the time of writing, first on Artificial Analysis's Intelligence Index. It is pitched at long-loop agentic work: the model you point at a task that runs for an hour and expect to still be on track at the end. Anthropic published agentic and coding evals at launch rather than the classic academic set, which is increasingly the pattern across labs.
Solthar DNA rating, out of 5
- reasoning: 5
- coding: 5
- writing: 4.5
- vision: 4
- speed: 3.5
- value: 4
- context: 5
- tools: 5
Verdict
The default for agentic coding and anything that has to stay coherent for a long run.