NEWantirez's ds4 Runs DeepSeek V4 Flash at 34.4 Tokens/s on an M5 Max, 14.4 on a DGX Spark
Reference rows for antirez's ds4 engine show a 128GB M5 Max generating DeepSeek V4 Flash over twice as fast as a DGX Spark, which prefills faster.
Seung Jung·
5m