Kimi K3.1 Pro leaked doc shows Gated DeepSeek Sparse Attention for long context
zephyr_z9 · x · 2026-11-04
A screenshot shared on X reveals details from Kimi K3.1 Pro's technical documentation: compared with Kimi K3, K3.1 Pro replaces full-attention Gated MLA layers with Gated DeepSeek Sparse Attention (Gated DSA) layers for more efficient long-context modeling.
The eval section notes that unless otherwise specified, all results use reasoning effort max, temperature = 1.0, and top-p = 0.95. The poster's "Whale victory" framing suggests Moonshot AI is borrowing DeepSeek's architecture to good effect.
Related event: Moonshot AI Open-Sources Kimi K3.1 Pro: 2.8T-Parameter MoE with 1M Context(17 posts)→
More from Models
- Retweet: Google's recurrent transformer could make sub-10B models viable soon — rickasaurus · 2026-11-05
- Kimi K3.1 Pro lands on Nebius Token Factory day 0, coding close to Opus 5.5 — HowDevelop · 2026-11-05
- GPT-6.2 Rumors Hold Up; Launch Could Come as Soon as Tomorrow — kimmonismus · 2026-11-05
- GPT-6.2 Could Launch as Soon as Friday in Astra, Sol, and Cyber Variants — kimmonismus · 2026-11-05
- Nathan Lambert launches free open model economy dashboard with paid analysis series — natolambert · 2026-11-05
- Underdog alert: Xiaomi MiMo keeps competing with top models on cost-adjusted benchmarks — VictorSanh · 2026-11-05