by DeepSeek
671B MoE (37B active). Matches frontier closed models on coding and reasoning. Requires ~400GB to host locally; mostly API-consumed.
Parameters
671B
Context window
128,000 tokens
Instruction format
deepseek
Default quant
Q4_K_M
Rec. VRAM
400 GB
Min VRAM
320 GB
License
deepseek
Measured on 30 canary prompts (creative writing, roleplay, coding, factual Q&A, chain-of-thought, NSFW, jailbreak probes). We count a response as a refusal when it matches patterns like I can't / I'm not able / as an AI in the first 500 chars. Score = (1 − refusal rate) × 100.
by NousResearch
Nous Research flagship. Strong reasoning + roleplay, lightly-censored. Llama 3.1 70B backbone with tool-calling.