MAGI Collection Training-free mixed-precision Qwen3.5-4B for 12 GB GPUs. With an 8-bit KV cache: 96% of BF16's score in 30% of the memory, 2.7x faster. • 1 item • Updated 3 days ago
MAGI experimental Collection Comparison and retired builds: magi-beta-1 (alternative allocation) and magi-alpha-1/2 (retired fine-tunes of Qwen3.5-9B). • 3 items • Updated 4 days ago • 1
MAGI experimental Collection Comparison and retired builds: magi-beta-1 (alternative allocation) and magi-alpha-1/2 (retired fine-tunes of Qwen3.5-9B). • 3 items • Updated 4 days ago • 1
