Post
krisman.dev
did:plc:o7dqe3ask7oyav2ddh75vbfz
Qwen 3.8 27B running (vllm/litellm) locally using Claude Code as the harness at the full 1M context on a single GPU. Every patch to make the integration with Claude Code seamless is included in the stack (and submitted upstream!) in the Github repo.
https://krisman.dev/blog/claude-code-local-qwen-blackwell
2026-08-21T18:58:34.666Z