#open source LLM

'AI2 releases Olmo 3, an open 7B and 32B LLM family built on the Dolma 3 data suite with a 65,536 token context and fully transparent training pipeline. Variants include reasoning, instruction-following, and RL-ready models for reproducible research.'

'MiniMax-M2 is a compact MoE released with open weights on Hugging Face, optimized for coding and agentic tool use with about 10B active parameters per token and claims of ~2x speed at ~8% of Claude Sonnet cost.'

DeepSeek-TNG introduces R1T2 Chimera, a new Assembly-of-Experts LLM that delivers twice the speed of R1-0528 and improved reasoning, available now under MIT license.