Tag
DeepSeek V4 Pro approaches frontier-level performance. Google, Mistral, and Alibaba ship under Apache 2.0. Ollama hits 52 million monthly downloads.
DeepSeek V4 matches Claude Opus on coding at 7x lower cost under MIT license. NVIDIA's Nemotron 3 brings hybrid Mamba-Transformer MoE to the open. Google's TurboQuant cuts KV cache memory by 6x with no retraining.
Google, Alibaba, Meta, Mistral, OpenAI, and Zhipu all ship competitive open-weight models under permissive licenses. The battleground shifts from benchmarks to inference speed on your actual GPU.
Mistral Small 4's 119B MoE unifies reasoning, vision, and coding—but needs datacenter hardware. Qwen 3.5 35B-A3B remains the consumer GPU king at 112 t/s.
Mistral releases a 4B parameter text-to-speech model that clones voices from 3 seconds of audio, runs locally on 16GB GPUs, and beats ElevenLabs in human evaluations.
Mistral drops a 119B MoE model under Apache 2.0, DeepSeek V4 emerges from stealth, and dual RTX 5090 setups are matching H100 on 70B inference. This week changed the game.
March 2026's open-source AI highlights: Zhipu's GLM-5 rivals GPT-5, OpenAI finally goes open, and the Linux Foundation creates a home for AI agents.
This week's biggest open-source AI developments: Alibaba's efficient new model outperforms its massive predecessor, Mistral releases a 675B frontier model under permissive license, and local inference adoption accelerates
