Unsloth Dynamic v3.1 GGUFs boost Qwen3.8-Flash-Next accuracy 10-15% over GSQ-RCO
danielhanchen · x · 2026-11-05
Unsloth released new Unsloth Dynamic v3.1 GGUF quants for Qwen3.8-Flash-Next, beating GSQ-RCO by 10-15% accuracy on DeepSWE, Terminal Bench, Aider Polyglot and more.
- New engram-switching guide covers BF16 (102.4 GB, RMSE 0), Q80 (54.4 GB, 0.0053), Q51 (38.4 GB, 0.0376), IQ4NL (28.8 GB, 0.0757)
- Builds on their DeepSeek-V4.1-Flash UD3.1 work, which improved accuracy 10-20% over alternatives
- Benchmarks: 256K context for agentic evals, 128K otherwise; temperature 1.0, topp 0.95, topk 20, no penalties; over 26B tokens benchmarked
- Runs on 32GB+ RAM setups; benchmarking/grading/infra moved to a new repo
Related event: Unsloth Releases Dynamic v3.1 Quantization for Qwen3.8-Flash-Next(4 posts)→
More from Infra
- AWS shows how to plug calibrated decision models into Bedrock AgentCore Evaluations — AWS ML Blog · 2026-11-05
- LFM3-Omni-2B ported to run fully offline on iPhone at ~80 tok/s with custom WebGPU kernels — JosephJacks_ · 2026-11-05
- Orbital Data Centers Launch to Sun-Synchronous Orbit, Memed as Elves Sailing West — rand_longevity · 2026-11-05
- Vietnam's Lâm Đồng to bring Starlink to mountain areas where poor signal hampers rescue — elonmusk · 2026-11-05
- SpaceX partners with Qualcomm for nationwide Starlink Mobile 5G network from 2027 — Polymarket · 2026-11-05
- Stopping runaway agent spend before the bill: pre-call guardrails and three gaps found the hard way — MaverikSh · 2026-11-05