Xiaomi releases MiMo-v2.5 Family weights with strong coding and agent benchmarks

 (firethering.com)

19 points | by steveharing1 4 hours ago

6 comments

  • swiftcoder 1 hour ago
    Has anyone run the numbers on Xiaomi's token subscription prices? The subscriptions are all priced in terms of "credits", and its not clear how credits map to tokens
  • steveharing1 2 hours ago
    It was released around a week ago but Xiaomi Open Sourced it Now & its MIT Licensed and available om HuggingFace
  • esperent 1 hour ago
    I thought I had a high tolerance for slop but this is literally painful to read.

    Here's a slightly better source that looks like a human at least reviewed it before publishing.

    https://venturebeat.com/ai/open-source-xiaomi-mimo-v2-5-and-...

  • landtuna 1 hour ago
    Ugh, the slop. "That's not a benchmark row. That's a story."
  • Alifatisk 1 hour ago
    I’ve been curious about MiMo models, they score so high on wide range of benchmarks, yet, not much is talked about it. At least in my bubble. Is it a serious contender? What is the model for?
    [-]
  • spwa4 1 hour ago
    TLDR: 2 models, 1 trillion parameters total, 42M active, 1M (MiMo-V2.5-Pro) or 256K (MiMo-V2.5-Pro-Base) context length, available on huggingface here:

    https://huggingface.co/XiaomiMiMo/MiMo-V2.5-Pro