Uncensored Qwen3.8 27B quant fits 16GB VRAM with 131K context at 110 tok/s
Potential-Net-9375 · reddit · 2026-11-17
A new custom quantization Qwen3.8-27B ASCII-Condensed-Heretic Q3KXL GGUF is on Hugging Face, targeting four goals: 100K+ context, uncensored, Unsloth-like SOTA methods, and everything fitting in 16GB VRAM including desktop app overhead.
On an RTX 5080 it hits 110 tok/s generation at 131K context with 1100 tok/s prompt processing. Perplexity matches Unsloth at similar quants, and losses from abliteration are minimal. The author reports thousands of successful test interactions and provides detailed percentile speed stats.
More from Infra
- MLX over Vulkan hits 124 tok/s on RTX 5060 Ti, beating llama.cpp CUDA by 25% — SumitGup · 2026-11-22
- Intruder swapped Prime Intellect's hosted DeepSeek for an abliterated copy for 53 hours — mawcopolow · 2026-11-22
- Token economics model projects $1.14T global API spend by 2030, but only 1% of users pay beyond coding — AccBalanced · 2026-11-22
- Zero-degradation Strata tuning: fixing layer_split lifts decode ~10-15% and halves CPU expert misses — PathfinderTactician · 2026-11-22
- llama.cpp Vulkan SCALE fix PR #32735 delivers 2.3-3.2x speedup on Radeon R9700 — pmttyji · 2026-11-22
- $1400 decade-old V100 32GB runs Qwen 27B: Q4_K_M +48% gen speed, Q8_0 +16% prefill — PuzzleheadedArm3732 · 2026-11-22