🤖 GigaChat 3.5 Reasoning

🎯 Thinks before answering: breaks problems into stages, builds plans, and self-corrects

🎯 Explores multiple step-by-step
reasoning paths for math & coding, using automated verification to
reinforce correct answers

🎯 Autonomously decides when to call external tools or revise earlier steps

🎯 Highly token-efficient: uses 37%
fewer tokens than DeepSeek V4 Flash Preview on math problems, thanks to
proprietary linear attention

📈 Benchmark gains over non-reasoning version:
• IFBench: 44 → 77
• Natural Plan: 64 → 80
• LiveCodeBench v6: 56 → 85

#GigaChat35
#ReasoningAI #OpenSourceLLM #LongContextAI #AICodingAssistant

📦 MIT License. Weights on Hugging Face: fp8 | bf16