ai-sage has released GigaChat 3.5 Ultra, a 432B parameter Mixture-of-Experts model designed for multilingual tasks, reasoning, and code generation. This new version is approximately 40% more compact than its predecessor, GigaChat 3.1 Ultra, while offering improved performance in coding and mathematical scenarios. GigaChat 3.5 Ultra features a hybrid attention architecture combining Multi-head Latent Attention with GatedDeltaNet, and incorporates Gated Normalization for better signal control and Multi-Token Prediction for faster generation. AI
IMPACT This release offers a more compact yet powerful model for multilingual tasks and agentic workloads, potentially improving inference efficiency.
RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
Read on Hugging Face Trending Models →
- ai-sage/GigaChat3.5-432B-A28B
- GatedDeltaNet
- Gated Normalization (GatedNorm)
- GigaChat
- GigaChat 3.1 Ultra
- GigaChat 3.5 Ultra
- GigaChat Ultra 3.0
- Habr
- Mixture of Experts (MoE)
- Multi-head Latent Attention (MLA)
- Multi-Token Prediction (MTP)
- RMSNorm
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
