Anthropic launches Claude Opus 5, claiming near-Fable 5 performance at half the price
Anthropic released Claude Opus 5 as its new primary model for coding and agents, matching top-tier benchmarks while maintaining previous pricing.
Anthropic has released Claude Opus 5, positioning the new model as its core system for programming, long-running AI agents, and professional tasks. According to the company's system card, the model approaches the performance of Fable 5 at half its API price.
The company highlights key improvements in agentic workflows. Claude Opus 5 reportedly checks its own work more thoroughly, traces root causes of errors more effectively, and continues troubleshooting past the first working solution. Pricing holds steady at $5 per million input tokens and $25 per million output tokens, matching the rate of Opus 4.8. A fast mode runs roughly 2.5 times faster at double the cost.
In benchmark results published in the model's system card, Claude Opus 5 scored 79.2% on SWE-bench Pro, nearly tying Fable 5 at 80%. On FrontierBench v0.1, it rose to 44.4% from the 18.7% posted by Opus 4.8, and scored 30.16% on ARC-AGI-3 compared to 7.78% for GPT-5.6 Sol. However, on DeepSWE, the model still trails GPT-5.6 Sol at 68.8% versus 72.7%. Most tests were conducted by Anthropic using adaptive thinking at maximum effort averaged over five runs.
Source: habr
Comments
Help shape the next version: Add context or suggest a correction. AI review can add points toward a rewrite. Reviews and updates may take time; a full meter does not guarantee a new version.