Choose your path

Use DeepSeek-V4-Flash for fast text automation. Choose GLM-5.3-Flash for multimodal agents & private hosting. Both models are MIT-licensed.

Set up DeepSeek Harness on your machine, configure an API key and review installation requirements. Understand local orchestration versus local inference.
DeepSeek V4 Flash 0731, with major agent benchmark gains, Responses API and Codex support, and new attention on DeepSeek Harness.