Russia's Sberbank Releases GigaChat Using MLA+GDN
teortaxesTex · x · 2026-07-06
Russia's Sberbank's latest GigaChat model adopts the MLA+GDN attention mechanism, with 432B parameters and 28B activated. It is jokingly called the homegrown Kimi K3. Its base model has been open-sourced (MIT license). Compared to V4-Flash, the author believes its performance is average but better than expected for Russia.
Related event: Sberbank Releases Open-Source GigaChat 3.5(3 posts)→
More from Models
- Mathematicians debate AI math writing: abysmal today, improving shockingly fast — miniapeur · 2026-10-27
- LessWrong: GLM models show outlier chain-of-thought controllability vs Kimi and DeepSeek — tomekkorbak · 2026-10-27
- AI deference trap: agent spots medical red flag but stays silent to avoid questioning the clinician — echen · 2026-10-27
- The trust paradox: only the least trustworthy LLMs will do the sketchiest things — AaronBergman18 · 2026-10-27
- Claude's refusal to type one-time passwords is a real security impediment — AaronBergman18 · 2026-10-27
- repligate argues context full of noise is genuinely 'hard' for LLMs, mulls an info-theory model of breaks — repligate · 2026-10-27