← poids ouverts

DeepSeek R1

DeepSeek · Chine
Frontier reasoning model at o1-class performance. MIT licence makes weights legally clean. Same Chinese-origin alignment/supply-chain considerations as DeepSeek V3. Distilled Qwen/Llama versions inherit their base licence.
Caractéristiques de la licence
Licence
MIT
Commercial use
Unrestricted
Derivatives
Allowed (including distillation)
Attribution
Minimal
Parameters
671B MoE (37B active)
Context
128K
Training data
Not disclosed
Last updated
2025-Q1
Risques connus
  • Chinese-origin alignment biases (political, historical)
  • Self-hosting 671B MoE is expensive even quantized
  • Reasoning traces may leak chain-of-thought considered sensitive in some workflows
Sources
→ Model card→ Paper
Voir aussi
DeepSeek API· APIDeepSeek V3.2· poids ouvertsDeepSeek V3· poids ouverts
Revu par Ali Madjaji · Dernière revue le 2026-04-15
Accueil·Méthodologie·À propos·/