xAI launched Grok 4.7, its latest reasoning model targeted at complex software engineering, long-horizon repo manipulation, and office tasks, while holding API pricing steady at $2 per million input tokens and $6 per million output tokens. Meanwhile, Xiaomi released MiMo-V2.5, an open-source model suite featuring a 1M context window and extensive reinforcement learning optimizations designed for agentic workflows.

We’ll talk about:

• Upgraded base model scoring 46.3% on CursorBench 4.0 and 71.0% on DeepSWE v1.1 while maintaining developer API rates.
• Release of MiMo-V2.5-Pro and general multimodal variants with full model weights, training reports, and agent evaluation frameworks.
• Meta’s standalone AI application outpacing early adoption rates across North America despite Amazon access restric ...