In-place refusal-direction edits of Qwen3.8-Flash-Next NVFP4. No tensor changes format — the edit is applied on the quantisation grid itself.
👋 Open to Work
Flare
lflare
AI & ML interests
None yet
Recent Activity
updated a model 2 days ago
lflare/ThinkingCap-Qwen3.6-27B-FP8 updated a model 2 days ago
lflare/Tess-4-27B-NVFP4A16 updated a model 2 days ago
lflare/Tess-4-27B-FP8Organizations
None yet
FP8 & NVFP4 quants of Qwen/Qwen3.8-27B. Reserved ahead of the upstream release — weights land as soon as the base model ships.
FP8 & NVFP4 quants of migtissera/Tess-4-27B, a Qwen3.6-27B reasoning fine-tune. KLD is measured against bf16 Tess, so it isolates quant loss.
FP8 & NVFP4 quants of Qwen/Qwen3.6-27B, plus the heretic-v2 uncensored line. Vision tower, SSM block and MTP head kept in bf16.
-
lflare/Qwen3.6-27B-NVFP4A16
Image-Text-to-Text • 28B • Updated • 57 • 1 -
lflare/Qwen3.6-27B-uncensored-heretic-v2-mtp
Image-Text-to-Text • 28B • Updated • 32 • 4 -
lflare/Qwen3.6-27B-uncensored-heretic-v2-mtp-FP8
Image-Text-to-Text • 28B • Updated • 55 • 5 -
lflare/Qwen3.6-27B-uncensored-heretic-v2-mtp-NVFP4A16
Image-Text-to-Text • 28B • Updated • 67 • 2
Two primitive-ai Qwen3.8-Flash-Next checkpoints x three quantised PLE tables. BF16 n-gram table dropped; one repo id per configuration.
FP8 & NVFP4 quants of the 256-expert Ornith-1.0-35B MoE. Routers, SSM block and vision tower stay bf16. Serve in vLLM or SGLang.
FP8 quants of bottlecapai/ThinkingCap-Qwen3.6-27B. bf16 SSM/attention layout, and unlike the upstream FP8 it loads in transformers on Blackwell.
In-place refusal-direction edits of Qwen3.8-Flash-Next NVFP4. No tensor changes format — the edit is applied on the quantisation grid itself.
Two primitive-ai Qwen3.8-Flash-Next checkpoints x three quantised PLE tables. BF16 n-gram table dropped; one repo id per configuration.
FP8 & NVFP4 quants of Qwen/Qwen3.8-27B. Reserved ahead of the upstream release — weights land as soon as the base model ships.
FP8 & NVFP4 quants of the 256-expert Ornith-1.0-35B MoE. Routers, SSM block and vision tower stay bf16. Serve in vLLM or SGLang.
FP8 & NVFP4 quants of migtissera/Tess-4-27B, a Qwen3.6-27B reasoning fine-tune. KLD is measured against bf16 Tess, so it isolates quant loss.
FP8 quants of bottlecapai/ThinkingCap-Qwen3.6-27B. bf16 SSM/attention layout, and unlike the upstream FP8 it loads in transformers on Blackwell.
FP8 & NVFP4 quants of Qwen/Qwen3.6-27B, plus the heretic-v2 uncensored line. Vision tower, SSM block and MTP head kept in bf16.
-
lflare/Qwen3.6-27B-NVFP4A16
Image-Text-to-Text • 28B • Updated • 57 • 1 -
lflare/Qwen3.6-27B-uncensored-heretic-v2-mtp
Image-Text-to-Text • 28B • Updated • 32 • 4 -
lflare/Qwen3.6-27B-uncensored-heretic-v2-mtp-FP8
Image-Text-to-Text • 28B • Updated • 55 • 5 -
lflare/Qwen3.6-27B-uncensored-heretic-v2-mtp-NVFP4A16
Image-Text-to-Text • 28B • Updated • 67 • 2