fleetml/Qwen3.8-27B-Uncensored-NVFP4-RTX5090
Image-Text-to-Text • 16B • Updated • 3.84k • 15
RTX 5090 optimized ModelOpt NVFP4 target and target-matched BF16 DSpark drafter as one speculative decoding release.
Note ModelOpt NVFP4 target optimized for one RTX 5090.
Note Target-matched BF16 DSpark speculative drafter.