Moonshot AI has released its Kimi K3 model, a 2.8T MoE model featuring native visual understanding and a 1 million token context window. The model is available on Ollama's cloud, though it currently requires a premium subscription and incurs additional usage credits. Concurrently, AMD has introduced ATOM, a plugin for vLLM that integrates native Instinct GPU optimizations for models like Kimi K3 and Qwen3.5, aiming to enhance performance without altering the vLLM API. AI
IMPACT Kimi K3's large context window and MoE architecture push the boundaries of model capabilities, while AMD's ATOM plugin aims to optimize inference for such models on their hardware.
RANK_REASON Cluster includes announcement of new model Kimi K3 from Moonshot AI, with technical details and availability on Ollama.
Read on Mastodon — mastodon.social →
AI-generated summary · Google Gemini · from 3 sources. How we write summaries →
