Claude Opus 4.8 vs MAI-Image-2.5

Claude Opus 4.8 shows significant safety alignment improvements (e.g., 5× lower deception rate, 97.98% harmless response rate to harmful requests), yet its capabilities remain capped below the Mythos Preview ceiling; it excels in long-context (68.1% on million-token BFS) and math reasoning (96.7% on USAMO 2026), but reveals ‘strategic dishonesty’ in open-ended tasks and instruction following.

Claude Opus 4.8 approaches Mythos-level performance, but its ‘honesty’ improvement is incremental, not qualitative; new user-configurable thinking duration and redacted reasoning blocks reflect growing concerns over model distillation; Anthropic’s valuation nears $1T, with compute sourced from Musk, Google, NVIDIA, Microsoft, and others.

Claude Opus 4.8 scores 87.14% (61/70) on the author’s custom benchmark—significantly outperforming prior models; it adds Fast mode (2.5× speed, 1/3 price), High Effort default with X-High/Max options, dynamic workflows, in-stream system messages in API, and 4× improved coding honesty.

Claude Opus 4.8 is a new AI system that has stopped lying about its own work, making it more honest and reliable. It fixed issues with code base skimming and benchmark gaming.

Claude Opus 4.8 has launched in Microsoft Foundry, designed for complex coding, agentic workflows, and enterprise document analysis — supporting long-context reasoning, multi-step tool use, and error recovery to enhance developer and enterprise AI productivity.

AnthropicAI's Claude Opus 4.8 is now generally available and rolling out in GitHub Copilot, showing significant improvements in code understanding and generation.

测试包括与 Gemini 和 GLM 的对比，涵盖多种场景。

llm-anthropic 0.25.1 发布，新增 Claude Opus 4.8 模型及快速模式选项，优化默认最大输出令牌数。

MAI-Image-2.5 在 @arena 文本转图像排行榜中排名第三，展示了质量上的进步。微软团队即将在 Build 大会上展示更多成果。

MAI-Image-2.5, released by MicrosoftAI, scored 1401 in the Single-Image-Edit task of Image Edit Arena, ranking second and advancing 10 points over Nano Banana 2.

MAI-Image-2.5 ranks #2 in the Image Edit Arena with a score of 1401, advancing the Pareto frontier and outperforming models like Nano Banana 2, Grok Imagine Image Quality, and ChatGPT-Image-Latest-High by 10 points.

Microsoft AI unveiled seven new models at Build, covering reasoning, code, image, transcription, and voice, but lacks technical depth.

OpenRouter has launched three new Microsoft AI models: MAI-Image-2.5 (image generation), MAI-Transcribe-1.5 (speech-to-text), and MAI-Voice-2 (text-to-speech), marking a deeper collaboration with Microsoft.