Unsloth ships Qwen3.8-27B Dynamic V3.1 GGUFs with 5-10% accuracy gains
danielhanchen · reddit · 2026-11-12
Unsloth released Dynamic V3.1 GGUF quants for Qwen3.8-27B, claiming 5-10% better accuracy on DeepSWE, Terminal Bench and more vs. UD3, GSQ-RCO and Exllama-3, with all quants from 2- to 6-bit updated and better algorithms planned for Qwen4. Key details: agentic benchmarks ran at 256K context (128K for others), scores repeated 6-16x. Unsloth also disputes ISTA's GSQ-RCO claims: ISTA's reported 100% AIME25 accuracy recovery appears based on a single trial, while Unsloth's 14 trials show 93-96% recovery with its older quants actually better on average. Including full embedding and RAM/disk footprint, Exllama-v3 also fares worse on the accuracy-vs-size curve.
More from Models
- Leak Claims OpenAI Will Launch Entire GPT-6.2 Family Tomorrow to Counter Opus 5.5 and Fable 5.2 — heypearlai · 2026-11-12
- ARC-AGI-3 leaderboard: models score 90%+ but at $10K+ per task — haider1 · 2026-11-12
- LMArena Launches Agent Arena for Head-to-Head Agent Evaluation — arena · 2026-11-12
- LLMs lack hyperlocal municipal data, explains disputed Texas power outage debate — csuwildcat · 2026-11-12
- Claude API's per-message effort lets you switch reasoning levels without breaking prompt cache — lyraxana · 2026-11-12
- rasbt coins "Benchmaxxing": making the numbers look better (best optics) — infoxiao · 2026-11-12