Anyone who uses LLM for coding should definitely give DeepSeek 4.1 Flash and/or Qwen3.8 27B a try.
I’ve been actively using DeepSeek 4.1 Flash via OpenRouter for just under 14 days now, and I’m very impressed with the results—which cost only a fraction of what Codex does. 😃
Claude would have been even more expensive in this comparison.
I had everything double-checked by Codex and only very rarely found an error—nothing really serious, just some clunky behavior, etc.!
It’s CRAZY what’s now possible with Open-Weights models in this context. 🤯😁
(Directly translated by deepl.com)
#DeepSeek_4-1 #Flash #Coding #Codex #Claude #LLM #Open_Weights #Models #Qwen3_8-27B #AI

