Anthropic released Claude Opus 5, which replaces Claude Opus 4.8 as the flagship Opus-tier model while maintaining pricing at $5 per million input tokens and $25 per million output tokens (MarkTechPost). On the ARC-AGI-3 benchmark for novel problem-solving, Opus 5 scores 30.2 percent, nearly four times higher than GPT-5.6 Sol, positioning it to approach Claude Fable 5’s capabilities at half the token cost (The Decoder). Claude’s voice mode now runs on Opus and Sonnet models across all platforms with access to email, calendar, and messaging integrations (The Decoder).
Datalab released Marker v2, a rewritten three-mode document-parsing pipeline that scores 76.0 on olmOCR-bench and sustains 2.9 pages per second on a single B200 GPU, outperforming MinerU’s backend by over 5× in speed while beating Docling on both accuracy and throughput metrics (MarkTechPost). Separately, a tutorial documented how to construct self-evolving AI agents using the OpenSpace framework, integrating custom skills, the Model Context Protocol (MCP), and SQLite-backed agent lineage tracking for efficient, reusable agent systems (MarkTechPost).