Thread

MiMo-v2.6-Pro: Intelligence, Performance and Price Analysis

166 points · 68 comments · theanonymousone

dom96 · 1790062122 · focus · HN ↗
It is an impressive model. Agreed on most that is written on this page, with the exception of it being fast. I ran it on my own LLM benchmark suite[1] and it is faster than DeepSeek but still much slower than leading models. But it's pricing is where it really shines.
KillSwitch-Bench 1.0
```
  Claude Opus 5           66.9
  GPT-6 Astra             57.9
  Claude Fable 5.1        46.7
  MiMo-V2.6-Pro           38.8
  Muse Spark 1.3          36.5
```
1 - <a href="https://bench.killswitch-lang.org/" rel="nofollow">https://bench.killswitch-lang.org/
1. ricardobeat · 1790074378 · focus · HN ↗
  Speed seems to vary a lot with demand. Last night it was reaching 80+ tok/s
2. drittich · 1790080932 · focus · HN ↗
  Interesting - have you done MiMo-v2.6-flash?
  1. dom96 · 1790093013 · focus · HN ↗
    Not yet. But can do so if there is interest.
    tlarkworthy · 1790526780 · focus · HN ↗
    I am interested in flash, I think that would be my default model
3. conception · 1790082155 · focus · HN ↗
  That looks like you aren’t using the ultra speed endpoint.
  1. dom96 · 1790092984 · focus · HN ↗
    I’m using openrouter which I think is a fair representation of what the typical user will experience.
4. gandreani · 1790084868 · focus · HN ↗
  Why does Luna have a score of 0? I will say that in my limited experience, I use Luna and DSv4 Flash (haven't tried 4.1 yet) and Luna is wayyyy faster. They both output at the same speed but DS has an endless thought process

Open on Hacker News to reply ↗

Unofficial Hacker News client; not affiliated with Y Combinator.