Qwen 3.6 27B Runs at 262K Context
GrungeWerX · reddit · 2026-07-16
The author reports that Qwen 3.6 27B can stably run up to a 262K context length locally, remaining "decent and coherent" during a week of testing beyond 200K.
They explained the need for such a long context due to MCP tools and memory/summary features causing roughly 40–50K of session startup bloat. The author is currently addressing this via an MCP broker server and other optimizations, and plans to experiment with Yarn to see how much further the context can be pushed.
More from coding & agent
- Tenable and AWS launch a Black Hat build event for open-source security agents and MCP servers — Dave_Maynor · 2026-07-22
- Codex helps build Valdiluce, an open-world game with climbing, gliding and gondolas — Dimillian · 2026-07-22
- HeyGen adds a media-sourcing skill for coding agents with 75k images and 10k tracks — HeyGen · 2026-07-22
- Agent search bottlenecks are now about variance, not raw latency — rohanpaul_ai · 2026-07-22
- LangSmith adds tracing for Pipecat, LiveKit, OpenAI Realtime, and Gemini Live — LangChain · 2026-07-22
- An MCP server signs every AI agent tool call into a verifiable Merkle chain — Funky_Chicken_22 · 2026-07-22