Moonshot ships Kimi K3, a 2.8-trillion-parameter open-weight mixture-of-experts model with native vision
Moonshot released Kimi K3 on 16 July — a 2.8-trillion-parameter open-weight mixture-of-experts model with native vision and a 1-million-token context window. It pushes the open-weight ceiling to a scale that used to belong only to the largest closed labs, and it arrives as open weights close the gap on everyday work to single-digit percentage points.
The parameter count is a statement of intent. A 2.8-trillion-parameter open-weight model says the open labs are no longer content to trail the frontier on scale — they are matching the largest closed systems and shipping the weights. The mixture-of-experts design keeps inference tractable by activating only a fraction of those parameters per token, which is how a model this large stays deployable.
Native vision and a million-token context are the capability envelope, not add-ons. Shipping multimodality and long-context in the base open-weight release means the ecosystem of fine-tunes and products built on K3 inherits those capabilities for free — the compounding advantage that has made the leading open models into default foundations rather than alternatives.
The market context is the cost collapse. With open weights now within single-digit percentage points of premium models on everyday work at four-to-ten-times-lower cost, a frontier-scale open release like K3 sharpens the question every buyer faces: what exactly is the closed premium buying, and for how much longer.
LLM-Stats — AI updates today (August 2026) — latest model releases → · BuildFastWithAI — Best open source AI models July 2026 →