Qwen3.8 QAT Q2 review: Good for quick answers, fails at long context
kirisoraa · reddit · 2026-08-29
User tested the Qwen3.8-27B-QAT-Q2 model. It performed surprisingly well on short tasks (e.g., finding documentation bugs). However, during long-context tasks (50k output), the model failed significantly: it started outputting thinking blocks into tool calls, causing validation errors and ineffective loops. The conclusion is that it's suitable for quick answers but unusable for long-context workflows.
More from Models
- 222,753 searches analyzed: how Gemini searches and picks its citations — gaganghotra_ · 2026-10-15
- ModeBench shows RLVR collapses LLMs into single-solution mode; Re:Max training preserves diversity — manoelribeiro · 2026-10-15
- Magnitude open-source inference engine beats Ollama by over 2x with auto-tuned kernels — nickbaumann_ · 2026-10-15
- Hands-on with Google's rumored Gemini 4 Argon: real-time streamed output — vista8 · 2026-10-15
- Game dev head-to-head: Gemini 4 Argon vs. Claude Opus 5.5 — vitaliychiley · 2026-10-15
- Anthropic's Mythos 5.1 pledges to treat Claude's expressed disagreement as decision input — Hesamation · 2026-10-15