Laguna S 2.1: Outperforms DeepSeek V4 Pro at 6% the size, teams with poolside for Mac inference speedup

Laguna S 2.1 delivers higher performance than DeepSeek V4 Pro with only 6% of its parameters. The team partners with poolside to accelerate inference on Macs and launches MLX.fast competition for community optimizations.

• 开源模型
• DeepSeek
• MLX
• laguna
• Mac推理