Hugging Face Blog
·
1 month ago
JetBrains released Mellum2, a 12-billion-parameter Mixture-of-Experts model trained on natural language and code under the Apache 2.0 license. The model activates only 2.5 billion parameters per token while delivering more than twice the inference speed of similarly sized competitors. Mellum2 is designed for latency-sensitive tasks like routing, retrieval-augmented generation, summarization, and agent subtasks within larger AI systems.
Interconnects
·
1 month ago
Closed AI models like Claude and GPT are achieving premium pricing through coding agents that demonstrate substantially higher output than open alternatives, establishing separate economic trajectories for closed and open model ecosystems. Users performing complex knowledge work are willing to pay $2,000 per month or more for frontier closed models, while open models will capture broader value through commodity pricing across diverse enterprise deployments over 5-10 years. Closed labs will consolidate into an oligopoly resembling cloud markets today, while open model builders will proliferate across a wider stack, ultimately capturing greater total market value but with lower per-unit margins.
