vLLM keeps hitting 400 context errors with DeepSeek harness; llama.cpp runs 24h+ fine
cviperr33 · reddit · 2026-08-23
Running Qwen3.8-27b on vLLM with DeepSeek Harness (DSH), the user hits constant 400 errors: the model caps at 115K tokens, but 82K prompt + 32K requested output exceeds it. Raising context to 142K and setting 115K in DSH still fails to compress correctly, and Hermes crashes the server near compaction. Meanwhile llama.cpp runs the same model 24h+ without errors. The post includes the full vLLM launch flags (qwen3coder tool parser, prefix caching, speculative decoding), but the author refuses to cap message context at 32K since the model often thinks at length.
More from coding & agent
- Self Labs launches Agent Vault to let AI agents log in and pay with credit cards — EXM7777 · 2026-10-16
- Plane Agents: triggers, not prompts, are what turn chatbots into workflow agents — JosephJacks_ · 2026-10-16
- Claude eng lead Felix Rieseberg publishes tutorial on being more ambitious with AI — danshipper · 2026-10-16
- Qwen Code Desktop v0.25.2 ships managed-agent fixes and MCP config repair — github-actions[bot] · 2026-10-16
- NVIDIA makes 7 AI agent courses free, from agent basics to securing agents — ifioknkem · 2026-10-16
- Claude runs 9-hour task, then stalls waiting for Drive permission — juliey4 · 2026-10-16