Understand this story
On July 16, 2026, Moonshot AI released its flagship model Kimi K3, a 2.8-trillion-parameter multimodal model with a 1M-token context window and native vision capabilities, built on a MoE architecture that activates 16 of 896 experts per token. According to a Recode China AI report, K3 scored 1,679 points on the Frontend Code Arena, is priced at $3 per million input tokens and $15 per million output tokens, and open weights are scheduled to be made available on July 27. On July 27, Kimi.ai announced that Modal and Baseten had become Day 0 launch partners for K3; Modal trained a custom DFlash speculator for the K3 architecture, raising interactive speed from 100 to 460 tokens/s; on the same day, SGLang and Miles announced Day-0 support, with the former handling inference and the latter handling RL training. On July 28, Kimi.ai announced that K3 went live on DigitalOcean Serverless Inference. On July 29, Unsloth released a 1-bit quantized GGUF version of Kimi K3, shrinking its size from 1.56TB to 594GB, a 62% reduction, while retaining about 78.9% of accuracy, and it can run on a Mac Studio with 128GB of memory.
AI-generated from public reports
Latest development · Report
English translation is being prepared.Developments
5 confirmed developments
- Source:Unsloth
- Source:Kimi.ai
- Source:Kimi.ai
- Source:LMSYS Blog
- Source:English text is being prepared.
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Jul 29
unslothaiFirst-partySelected
English translation is being prepared.
Jul 28
kimi_moonshotFirst-party
English translation is being prepared.
Jul 27
kimi_moonshotFirst-party
English translation is being prepared.
kimi_moonshotFirst-party
English translation is being prepared.
modalFirst-partySelected
English translation is being prepared.
lmsys_blogFirst-partySelected
English translation is being prepared.
Jul 22
Interconnects
Open Model Roundup: Kimi K3, Qwen 3.8, the Distillation Controversy, and the Open-Closed Gap
The Interconnects podcast uses last week's release of Kimi K3 as the starting point for a quarterly roundup of open models, discussing its gap to the closed-source frontier, the continuing role of GLM 5.2, and progress from Chinese vendors such as Qwen, DeepSeek, and MiniMax. The episode also covers the distillation controversy, the U.S. open-model ecosystem, the distinction between frontier and near-frontier models, and cybersecurity arguments against bans.
Jul 20
interconnects
English translation is being prepared.
recode_china_aiFirst-partySelected
English translation is being prepared.
Jul 18
zhang_benita
English translation is being prepared.