MiMo-V2.5
Xiaomiwe adopt an FP8 mixed-precision framework similar to DeepSeek-V3
2 more from this model
Trained on a total of ~48T tokens using FP8 mixed precision.
Trained on a total of ~48T tokens using FP8 mixed precision.
specific method · filed under optimization
Training using mixed precision with FP8; the evidence does not specify a more particular precision recipe.
The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See training precision for the methods filed beside it.
Also called FP8 mixed precision.
4 quotes from 4 sources, one card per model, each opening with its strongest.
we adopt an FP8 mixed-precision framework similar to DeepSeek-V3
Trained on a total of ~48T tokens using FP8 mixed precision.
Trained on a total of ~48T tokens using FP8 mixed precision.
Trained on 27T tokens using FP8 mixed precision
Picked by hand, not extracted: where to read more, not evidence for anything on this page.