DeepSeek: X2 Speed for WASM with SIMD

Re:DeepSeek: X2 Speed for WASM with SIMD by fragmede on Tuesday January 28, 2025 @02:13PM (#42856618)

yeah:

    aider --model ollama_chat/deepseek-r1:32b

(or whatever)

[ Parent | on HN ]

Re:DeepSeek: X2 Speed for WASM with SIMD by sureglymop on Tuesday January 28, 2025 @04:01PM (#42857939)

This didn't work well for me, no changes are ever made but maybe it's because I'm just using the 14B model.

[ Parent | on HN ]

Re:DeepSeek: X2 Speed for WASM with SIMD by manmal on Tuesday January 28, 2025 @05:07PM (#42858700)

In case you are on a 32+GB Mac, you could try deepseek-r1-distill-qwen-32b-mlx in LM Studio. It’s just barely usable speed-wise, but gives useful results most of the time.

[ Parent | on HN ]