The Agent framework DeepSeek Harness, which was open sourced last night, received 20,000 stars in a few hours.
I installed 0.1.0 that night: the local webpage starts in 80 seconds, so you need to bring your own model key. Three tasks are given in one sentence: a resume generator, a mini-game that can be cleared, and its self-introduction page. 3/3 are all handed in.
197,000+ people have starred an open-source AI coding agent that connects to 75+ model providers
OpenCode sits between your codebase and the model Claude, GPT, Gemini, DeepSeek, OpenRouter and local Ollama models can all use the same project context, tools and permission rules
Tests of new models - of course the standard test - Chain Reaction Machine
Google/gemini-3.7-flash - $2.97 - x-ai/grok-4.6 - $2.76 - meta/muse-spark-1.2 - $0.21 - bytedance-seed/seed-2.0-code - $0.31 One note and observation - look what gemini-3.7-flash did????
DeepSeek V4 Flash succeeded in running locally on AMD Strix Halo (Ryzen AI Max+ 395 / 128GB)
The chat UI is about 15 t/s, and when I have a conversation it takes more than a minute (👇 that part is cut in the video) Lightweight Atomic Agent when using agents
You have reached the end of the archive
All of deepseek