Training GigaChat MAX
summary
Presentation describes GigaChat MAX, a large Russian-language LLM launched in 2025; it combines state-of-the-art distributed training technologies and curated high-quality datasets to achieve strong Russian-language performance. The talk covers the model’s architecture and the distributed training approaches used to scale learning and data curation.
AI-written summary · about
chapters (7)
- 0:00Introduction to GigaChat MAX
- 3:18Training GigaChat Lite
- 8:59Lite to Pro: architecture
- 15:45Technologies for GigaChat Pro
- 22:32Cluster issues and solutions
- 30:00Training GigaChat MAX
- 41:02Q&A
related talks
41mIntegrating GigaChat LLM into a virtual assistant: technical implementation
14mTechTalk "YandexGPT: How We Built It and What’s Next"
30mHow To Build Your Own Internal ChatGPT
35mHow We Launched YandexGPT
01h 01mHow ChatGPT Works
25mEmbeddings, Transformers, RLHF: Three key ideas to understand ChatGPT
22mIntroduction to Large Language Models (LLM) on Kubernetes
56mLarge Language Models for Python Specialists