Anthropic Releases Claude Opus 4.8 with Enhanced AI Capabilities
- Anthropic launched Claude Opus 4.8 six weeks after the previous version, maintaining the same pricing of $5 per million input tokens and $25 per million output tokens.
- Opus 4.8 improved its score on the SWE-bench Pro to a passing rate of 69.2%, outperforming OpenAI’s GPT-5.5 at 58.6% and Google’s Gemini at 54.2%.
- The model’s alignment scores are now comparable to Anthropic’s restricted frontier model, Claude Mythos Preview, with significantly lower rates of deceptive behavior than Opus 4.7.
Claude Opus 4.8 offers enhanced performance in software engineering and reasoning tasks without an increase in cost, making it a competitive option among AI models at its price point.
Despite its higher cost compared to Chinese models like DeepSeek V4 Pro, Anthropic emphasizes quality and safety as key differentiators for Opus 4.8. (Source)
