DeepSeek-V4-Flash

DeepSeek-V4-Flash 0731 is official: MIT weights, 284B MoE (13B active), 1M context, and agent benchmarks above V4-Pro-Preview at $0.14/$0.28 per 1M tokens.

Inkling

Mira Murati's Thinking Machines Lab shipped Inkling: 975B parameters, Apache 2.0, weights on Hugging Face on day one. The largest US open-weights model yet, with multimodal input and a thinking effort dial. Here is what it is, what it scores, and how to run or fine-tune it today.