Gemini 2.5 Pro Preview 05-06 (I/O edition) - the SOTA vision+coding model

**Gemini 2.5 Pro** has been updated with enhanced multimodal image-to-code capabilities and dominates the WebDev Arena Leaderboard, surpassing **Claude 3.7 Sonnet** in coding and other tasks. **Nvidia** released the **Llama-Nemotron** model family on Hugging Face, noted for efficient reasoning and inference. **Alibaba's Qwen3** models range from 0.6B to 235B parameters, including dense and MoE variants. **KerasRS** was released by **Franois Chollet** as a new recommender system library compatible with JAX, PyTorch, and TensorFlow, optimized for TPUs. These updates highlight advancements in coding, reasoning, and speech recognition models.