Re:MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per s...
by npn on Monday June 08, 2026 @11:57AM
(#48447050)
How?
[ Parent
| on HN ]
Re:MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per s...
by lostmsu on Monday June 08, 2026 @12:43PM
(#48447691)
They say they are using https://github.com/tile-ai/TileRT
- persistent CUDA kernel
- tiled processing with overlapping read/writes
- model designed with specific constraints in mind
Re:MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per s...
by aitchnyu on Monday June 08, 2026 @03:06PM
(#48449997)
Excuse me, do aliens live among us? 17 commits, 99% Python and multiplying the speed of GLM, Deepseek V4, MiMO 2.5?
Re:MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per s...
by zander_jiang on Tuesday June 09, 2026 @02:45AM
(#48457438)
tilert is closed source, the repo is just a python wrapper that invokes the binary.
Re:MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per s...
by 2001zhaozhao on Monday June 08, 2026 @02:09PM
(#48448932)
i wonder if it will be possible to hardcode a model with some kind of MTP-adjacent algorithm to use a smaller portion of it to generate most of the tokens but route to the real experts every once in a while to steer it towards good thinking directions. (Perhaps this is done only when it's generating its thinking block, and the training takes it into account)
Could result in very high efficiency and still good intelligence without having to resort to fundamental adjustments like going to a diffusion LLM
Re:MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per s...
by npn on Monday June 08, 2026 @02:34PM
(#48449428)
I doubt you can do that. MTP magic happens because for texts, we have a lot of low value fixed tokens that almost always get generated in the sequence (like punctuation, function words, language keywords etc). for most important ones (the entities, the content words, variables) you still need the full model.
so there is alwasy a maximum limit for how well MTP can do.