#Ad
#AI_Models
๐ฅ GigaChat 3.5 Reasoning [Open-Source]
โน๏ธ Overview:
New LLM that thinks before it answers. Breaks problems into stages, builds plans, checks results, and self-corrects using automated verification.
๐ Source:
Hugging Face fp8 | bf16
๐ Model Specs:
โช Built on GigaChat 3.5 Ultra with multiple step-by-step reasoning paths
โช Proprietary linear attention for efficient long contexts
โช Token-efficient: 37% fewer tokens than DeepSeek V4 Flash Preview
โช Benchmarks: IFBench 44โ77, Natural Plan 64โ80, LiveCodeBench v6 56โ85
โช MIT License
Post #2427
2.78K
- โค 2