User reports DeepSeek V4.1 Flash easily exceeds 150 tok/s

Josh Stevenson | RecursiveIntell @RecursiveIntellOriginally posted 2026.10.22

Open original on X

I'm seriously glad about @openai 20x to 10x thing more now. I realized even free ai is as good or better than chatgpt now and that opencode go $10 plan goes a long way with glm 5.3 flash at like 100+ tps vs @openai 15-25'ish tps. Deepseek v4.1 flash is easily 150+ tps. Then there's @ollama who give you $60 worth of inference for $20, no 5 hour or weekly limit and also high tps. My point is, you can get high quality intelligence these days for good prices and if you plan right, you don't need the frontier models to get your work done.