Kimi K3 Released: 2.8 Trillion Parameters
1littlecoder · youtube · 2026-07-16
The video introduces Kimi K3: a 2.8 trillion parameter model based on the Kimi Delta Attention hybrid linear attention mechanism, incorporating Attention Residuals.
Key points include:
- Native visual understanding capabilities
- 1M token context window
- Targeted at software engineering, knowledge work, and deep reasoning scenarios
- Evaluated as having overall intelligence performance second only to Claude Fable 5 and GPT-5.6 Sol
- Model weights will be released within a few days, with a technical report detailing the architecture, training, and evaluation
The video also mentions that it can be accessed via official documentation, pricing is available, and a brief comparison test with Sol 5.6 is included.
Related event: Kimi K3 Debuts Strong, Narrowing the Open-Weight Gap(184 posts)→
More from Models
- Kimi K2.8 Preview rolls out: near-K3 coding performance, 1M context for all tiers — teortaxesTex · 2026-09-11
- Looking for a classifier of software engineering task shapes to pick models per task — StewartalsopIII · 2026-09-11
- DeepSeek V4 Pro API to continue after Sept 2026, billing unchanged — teortaxesTex · 2026-09-11
- DeepSeek V4.1 Flash Hits 98% of GPT-6 Astra's Score at 1.4% of the Cost in Third-Party Benchmark — ayushtweetshere · 2026-09-11
- TheZvi Polls: Has Your Coding Model Choice Changed Since Fable 5.1 and Astra? — TheZvi · 2026-09-11
- antirez Weighs In on Anthropic Banning Minors From Using Claude — antirez · 2026-09-11