DIY Heretic quant of Qwen 3.8 Flash Next ships as 4-bit Unsloth-style GGUF
jjusko20 · reddit · 2026-10-22
Reddit user jjusko20 released Swift-1.5-Qwen-3.8-Flash-Next-Heretic-UD-IQ4XS, filling a gap since only GSQ-RCO uncensored editions existed before.
Key points:
- Full-precision Swift 1.5 processed with the Heretic refusal-removal method, then quantized to 4-bit IQ4XS GGUF
- Tensor profile extracted from the UD-IQ4XS GGUF of Qwen 3.8 Flash Next, replicating Unsloth dynamic quants; Strata comes preconfigured for this build
- imatrix calibration covered math, code, English, general instruct, and tools, generated against a small custom dataset
Available on Hugging Face; the author takes requests for new formats.
More from Infra
- pi ships rec inference engine, runs locally on M5 Max in two lines — _rockt · 2026-10-22
- Researcher warns GitLab flaw is 'really bad,' urges users to patch now — basedjensen · 2026-10-22
- Meta Beats With $64.61B Revenue, Raises 2026 CapEx Guidance to $135-145B — RihardJarc · 2026-10-22
- PyTorch Con Demo Theater: 13 talks including Wan 2.2 video generation sped up 53.6x — PyTorch · 2026-10-22
- TorchTitan adds 5D parallelism for DeepSeek-V4, Qwen3.5, Kimi K3; Muon optimizer 79% faster — PyTorch · 2026-10-22
- JNTC cuts TGV glass substrate copper filling from a day to minutes, targets 2027 mass production — toptickcrypto · 2026-10-22