[perfmode]

How’s the token throughput / response time?

[simonw]

Healthy!

  prefill: 30.91 t/s, generation: 29.58 t/s

From https://gist.github.com/simonw/31127f9025845c4c9b10c3e0d8612...

[antirez]

Prefill is 400 t/s in that hardware. Just if the prompt is very short you can't see the real speed and it will default to single token context processing.

[simonw]

Hah, that's my fault for just using "Generate an SVG of a pelican riding a bicycle" as my test prompt!

[xienze]

I don't want to be a jerk but 31t/s prefill is basically unusable in an agentic situation. A mere 10k in context and you're sitting there for 5+ minutes before the first token is generated.

[fgfarben]

That prefill number isn't right. M4 Max hits 200-300: https://github.com/antirez/ds4/blob/main/speed-bench/m4_max_...

[hadlock]

M5 studio is gonna sell like hot cakes

[throwdbaaway]

Hah, that's because the prompt itself was only about 30 tokens. We need a much bigger prompt to properly test PP.

[aiscoming]

if it's just the coding agent system prompt and tools, you can cache that

[xienze]

Yeah the problem is that's just the start of the context. There's, you know, all the tool call results and file reads and stuff.

[embedding-shape]

Comparison with a RTX Pro 6000, with DeepSeek-V4-Flash-IQ2XXS-w2Q2K-AProjQ8-SExpQ8-OutQ8-chat-v2-imatrix.gguf:

prefill: 121.76 t/s, generation: 47.85 t/s

Main target seems to be Apple's Metal, so makes sense. Might be fun to see how fast one could make it go though :) The model seems really good too, even though it's in IQ2.

[incidentist]

Someone is working on a fork that is optimized for M5, might be worth a look: https://github.com/Swival/ds4-m5

[rtpg]

what are token speeds like for frontier models, if that gives a rough idea of how much "slower" slow is?