stevibe(@stevibe):Google dropped MTP versions of Gemma4. Ran them on my DGX Spark. The 31B dense model went from 3.94 → 8.91 tok/s. That's +126%. Full results: [26B A4B] > 25.24 → 31.69 tok/s (+25.6%) > TTFT 755 → 332ms (-56%) [31B] > 3.94 → 8.91 tok/s (+126%) > TTFT 599 → 378ms (-37%) If you're not running MTP, you're leaving free perf on the table.

stevibe

@stevibe

LLM. Local AI addict. Building @BenchLocalAI Builds things nobody asked for. Benchmarks things for fun.

가입 July 2009

1.3K 팔로잉 중 27.7K 팬

stevibe@stevibe

2026.05.07 09:36

Google dropped MTP versions of Gemma4. Ran them on my DGX Spark. The 31B dense model went from 3.94 → 8.91 tok/s. That's +126%. Full results: [26B A4B] > 25.24 → 31.69 tok/s (+25.6%) > TTFT 755 → 332ms (-56%) [31B] > 3.94 → 8.91 tok/s (+126%) > TTFT 599 → 378ms (-37%) If you're not running MTP, you're leaving free perf on the table.