Swift efficient reasoning + HyperQwen serving: three checkpoints compared on RTX 3090 task time, token use, throughput and quality.
🔄 In a Training Loop
David Hauser
daavidhauser
P(doom) <0.1%
AI & ML interests
ML 4 Science, LLMs and more.
Recent Activity
updated a collection 13 days ago
Swift for HyperQwen — RTX 3090 benchmarks updated a collection 13 days ago
Swift for HyperQwen — RTX 3090 benchmarks updated a model 13 days ago
daavidhauser/Swift-1.5-Qwen3.8-27B-W4A16-HyperQwenOrganizations
None yet