Key Info

OpenCode announced that DeepSeek V4.1 Flash and GLM 5.3 are now priced 60% lower than Fireworks and Baseten, with the company claiming it handles more volume than those providers, all hosted on its own GPUs.

Highlights

  • DeepSeek V4.1 Flash and GLM 5.3 are offered at 60% below Fireworks/Baseten pricing.
  • OpenCode says it serves more volume than those competitors.
  • All inference runs on OpenCode's own GPUs rather than third-party capacity.
  • The author says bad experiences with inference providers—commit lock-ins, idle-GPU billing, aggressive sales, and a discount-for-data offer tied to an unmanaged Alibaba endpoint—led OpenCode to build its own inference stack.