2026-10-30 · America/Los_Angeles · 社区动态 · #14

DeepSeek V4.1 Flash on Colab. Uncensored. 118.7 tok/s.

The uncensored 249 GB model as your own OpenAI-compatible endpoint, at over 100 tokens a second. One command, one Google Colab G4, no server of your own. Runs on a Google AI Pro subscription: 200 compute units every month. https://github.com/leonxger/deepseek-colab

热度 45.2 / 100；排名与评分保留该期记录。