benchmarkIMAGE
MiMo RL run lifts pro model to 72.57 on DeepSWE, still below top score of 74
The author says MiMo's RL training has finished, with the pro model rising from 58.41 to 72.57 on DeepSWE, still under the top score of 74 held by Astra, Gemini 3.8 Flash and Opus 5.
wh
@nrehiew_Archived 2026.09.21Originally posted 2026.09.21
wh @nrehiew_Originally posted 2026.09.21
Looks like the MiMo RL runs have completed. The pro model went from 58.41 to 72.57 on DeepSWE. For context, the highest score on DeepSWE is 74 by Astra, Gemini 3.8 Flash and Opus 5
x.com/nrehiew_/status/2101851063364333824Archived 2026.09.21 10:19