Xiaomi MiMo v2.6

(mimo.xiaomi.com)

248 points | by volf_ 2 hours ago ago

99 comments

  • rao-v an hour ago

    I know we have strong views on what a truly open model is (open weights, open training data, open training code etc.) but I really like how transparent they’ve been about the training of this model.

    The realtime dashboard they shared during training (https://mimo.xiaomi.com/rl/) was an incredible learning and teaching tool for me, and they’ve been unusually comprehensive in sharing details about their methodology (check out that tech report - it's got lots of clever behind the scene tricks like Google or Deepseek writeups) and benchmark scores (even the stuff they didn’t do well on).

    If you’re releasing an open model going forward, please consider offering the community more of this transparency!

    • ignoramous a minute ago

      [delayed]

    • earthnail an hour ago

      Thanks so much for sharing this. As someone who mostly watches from the sideline, can you share what you can see in this dashboard that someone like me can't see? Is it the metrics themselves that they measure (the metrics tab is absurdly detailed), something in the notices, or something else I missed?

      • rao-v 8 minutes ago

        I might turn this into a blogpost if folks are interested, but my god there is so much clever info in that dashboard.

        Here is one really neat bit:

        A cutting edge training idea (for agents, it's been used elsewhere for ages) is on-policy RL, basically, it's not enough to say "here is an end to end agentic sequence (including tool calls etc.) that is perfect" you want to say "here is a sequence you might actually have generated that turns out to be correct".

        Basically, it's more training efficient for models to improve with small tweaks to what they already do than from some perfect oracular answer

        (if you've ever tried to teach humans new skills, you’ve probably noticed this too!)

        When you do that, you care about how far the model you are updating (improving) has deviated from the one being used to generate rollouts (agentic rollouts for hard problems can take hours with lots of tool calls, so you can't keep redeploying every slight improvement).

        Lo and behold, the dashboard literally has:

        partial/avg_staleness (likely the measure of how many micro iterations the "generate answers" model is behind the "improving based on the occasional right answer" model)

        train_infer_diff/new_infer/kl (a more direct KL divergence based way of measuring how differently the two models generate tokens)

        How cool is that?!

      • tancop 18 minutes ago

        The best thing they did is being open about all the setbacks they had to deal with. They logged every restart with a reason, talked about dropping a cyber dataset after it degraded coding benchmarks. Also published real time training loss, benchmark scores after every checkpoint and running cost estimates.

        Really the only thing missing was dataset descriptions, the dashboard only had random IDs like "dataset-zrso". I guess it's their lawyers fault.

      • verdverm 36 minutes ago

        the existence, who else has a live dashboard for the RL late-training?

    • MangoCoffee 8 minutes ago

      maybe this is why Dario want to slow down AI development and all the big AI labs in the USA is singing the same song.

      whey they all singing the same tune. it make me question what is their real motives.

      they are afraid of Chinese good enough LLM model killing their margin. we already have story about US companies switch some task to use cheaper Chinese model hosted on Neoclouds.

  • simonw 41 minutes ago
    • phainopepla2 33 minutes ago

      I think we can say pretty confidently they aren't pelican-bench-maxxing

    • brcmthrowaway 33 minutes ago

      Just me, or do these look bad?

      Qwen3.8-27b pelican was amazing on Mac.

      https://www.nudgehost.com/dpjn3uwe

      • idiotsecant 30 minutes ago

        Looking terrible isn't nessesarily a bad thing. The pelican is heavily pre trained now. Having a crappy pelican means you didn't try to juke the stats.

        • broodbucket 13 minutes ago

          Apologies for not taking the time to find it, but there was a post that tried to determine if the pelican was benchmaxxed across a bunch of models by comparing it to other SVGs, and found that it wasn't at all.

    • handfuloflight 30 minutes ago

      How does this translate to coding performance, which is what most of HN cares about (...I assume)?

      • simonw 28 minutes ago

        It means they're good at writing SVGs, in particular SVGs of animals riding modes of transport!

      • lanyard-textile 24 minutes ago

        I only visit HN for the pelicans, personally.

        • lukewrites 13 minutes ago

          Yeah, I thought the "N" was for Nest

    • Imanari 33 minutes ago

      ish… at least we can be sure they don’t benchmaxx the pelicans lol

  • lwansbrough an hour ago

    Anyone else more excited about Chinese models than American models these days? Big thing for me is affordability.

    • tacomagick an hour ago

      Absolutely! Chinese models are both cheaper and more capable in many cases, compared to the American models and their makers continuously fumbling or reducing model capability with each update. Deepseek decreased costs when they released Flash 4.1 you would not see any American company do this, in reverse they would try charge you more.

      • user43928 38 minutes ago

        OpenAI decreased prices with the 5.6 model family.

        And later they further cut Sol and Terra pricing by 20% (maybe only in the API) and Luna by 80%.

        In fact Luna still outperformed DeepSeek Flash 4.1 in cost per task on Artificial Analysis when I last checked.

        However, Luna is slightly less intelligent. I have a feeling that it's pretty dumb and prone to hallucination unless running at xhigh or max effort, where it somehow manages to work quite well.

        I did not personally test the open weight models beyond the old Qwen 3.6 27B, which produced unusably bad results for me.

        The competition is great, and I hope Chinese models will continue to force leading US labs to offer models at a low price point.

        That said, I don't think the Chinese labs have anything over OpenAI and Anthropic when it comes to capability or efficiency - I have no reason not to believe the US labs have even lower cost to serve the models.

        • tacomagick 27 minutes ago

          OpenAI had to cut costs because of Anthropic. I also do not trust the benchmarks when it comes to models anymore. I have tried both Claude and OpenAI models and while it is true that the 5.6 series is smarter than Deepseek (at the time i tested it against 4.0) at that price it is still not worth it and sometimes randomly refuses to do tasks or stops midway etc.

          Do also remember China is this far in the AI race despite all chip restrictions from America. If they were in equal standards I truly think Chinese models would have long surpassed American ones. Also would like to remind how Anthropic CEO is being hostile and blaming Chinese models with distilling meanwhile their own models claimed to be Qwen¹ and their stance against open models is negative² and they still keep blaming China for it.

          1- https://news.ycombinator.com/item?id=48671252

          2-https://www.anthropic.com/news/position-open-weights-models

          • user43928 9 minutes ago

            Not sure about that.

            Given the difference in compute, it seems plausible.

            However, the researchers at the US labs are surely no less talented, and they have better access to hire talent globally.

            They too have to serve their models efficiently at a large scale, and with current capacity constraints this must be a top priority.

          • goosejuice 11 minutes ago

            > Also would like to remind how Anthropic CEO is being hostile and blaming Chinese models with distilling

            Why wouldn't he? If there really was 25,000 accounts breaking ToS any CEO would at minimum be upset. Evidence of Claude distilling qwen would be damning but that a) makes no sense b) doesn't exist afaik.

      • goosejuice 19 minutes ago

        > Deepseek decreased costs when they released Flash 4.1 you would not see any American company do this, in reverse they would try charge you more.

        OpenAI reduced prices and Anthropic increased weekly usage limits.

    • SyneRyder 29 minutes ago

      Yep, I'm trending in that direction, and I'm someone with Claude stickers all over my laptop. My main app dev work is still going to Claude, but everything else is going to China even at API rates now.

      One simple task: I needed an LLM to go through and clean up a few thousand page descriptions and titles in my personal search engine index, where the human web page authors had put in no effort sigh. I did a shoot out between Claude, Luna, GLM 5.3 Flash and Deepseek. Despite the high cost, Claude's descriptions were terrible, and even Opus warned me that the descriptions coming back from Haiku were "generalized, not accurate". I expected I would choose Luna because of price, and occasionally it did have wonderful descriptions (one captured emotion in a way no other model did). But in the end, the GLM 5.3 Flash descriptions were the easiest to read, they flow well while also being accurate & including necessary keywords, and being highly affordable. So it won out. It's a task that is nowhere near frontier, but a task where somehow China is better than frontier.

    • swingandamiss an hour ago

      No, because I'd rather not support our economic and military rivals.

      • lwansbrough an hour ago

        I'm Canadian so this sentiment has little value in 2026 unfortunately.

        • ActionHank 39 minutes ago

          Also, frankly, as a fellow Canadian it's pretty clear that the biggest "rival" the US has right now is itself. Just passed out in the corner puking on itself shouting about all the foreigners who won't talk to it.

        • tancop 9 minutes ago

          I'm from Europe and I hate America way more than China now. Used to be about equal but then Trump started extorting Ukraine, threatening their own allies and sending billions to Israel to help with a genocide. I think that exposed America for what it really is.

        • rayiner 21 minutes ago

          Canadians warming up to China makes me think of Germany becoming increasingly reliant on Russia in the 2010s.

          • bigyabai a minute ago

            Mind you, Canada was a reliable defense and trade partner before the US' arbitrary trade war and annexation threats.

      • Freedom2 an hour ago

        Agreed, and also because I support freedom of speech!

        • girvo 44 minutes ago

          Neither the US nor the Chinese companies are on your side then. They both censor, just different topics.

          But at least I can run Chinese models locally, and strip a lot of that censorship/refusal.

    • verdverm 34 minutes ago

      I have a contrarian opinion that China passing America in Ai is the Sputnik moment we need to leave the hubris behind and get our mojo back

      debatable if a turn around is possible before '29

  • stymaar an hour ago

    Flash[1]: 309B total / 15B activated parameters

    Pro [2]:, 1.02T total / 42B activated parameters

    [1]: https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL

    [2]: https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Pro-RL

    • verdverm an hour ago
      • gandreani an hour ago

        Those this mean they've fine-tuned this Qwen 3.5 9B on output from the V2.6 model?

        • mydreamof an hour ago

          It is a 9B agentic model developed by Xiaomi MiMo through supervised fine-tuning of Qwen3.5-9B on MiMo-generated data

    • verdverm an hour ago

      curious why the HF pill (on the right) always has inaccurate values

      • stymaar an hour ago

        I noticed the same, and I wonder as well.

        • verdverm an hour ago

          I suspect they are calculating something in the weights or config, I see it pretty consistently with quants

    • segmondy 27 minutes ago

      more like 500B in FP8

  • nemothekid an hour ago

    Looking at the frontend design examples; why do these models seem to love the "01 - UPPERCASE TEXT" motif. It's everywhere now (see https://try.cloudflare.com/, which has '01 · QUICK TUNNELS', but no "02" anywhere).

    • danvayn an hour ago

      My guess is that by function they break down frontend sections or components into pieces and I believe document things for themselves on some level, or purposely are verbose in this way. It is probably also shaped by users and existing web patterns. They probably get reinforced by models the more common they become.

    • sandblast an hour ago

      Nice catch!

    • pphysch 21 minutes ago

      The extraneous small-caps labels are one of the main idiosyncrasies of AI generated markup. I wonder how much of this is a "scaffolding" technique to help the model build stable designs. But was it reinforced in RLHF or an emergent behavior of the models?

  • user43928 an hour ago

    I don't trust any of the benchmarks where Opus 5 surpasses Astra or Fable 5.1.

    Maybe Terminal Bench 4.0 and ExploitGym are reasonable.

    Terminal Bench 4.0

      GPT 6 Astra             59.6
      Claude Fable 5.1        55.1
      Claude Opus 5           49.0
      MiMo-V2.6-Pro           34.9
      MiMo-V2.6-Flash         28.8
      DeepSeek V4.1 Flash     26.8
      MiMo-V2.5-Pro            1.5
    
    ExploitGym

      GPT 6 Astra             42.4
      Claude Fable 5.1        30.4
      Claude Opus 5           22.1
      MiMo-V2.6-Pro           17.8
      MiMo-V2.6-Flash          6.0
      MiMo-V2.5-Pro            0.1
    
    DeepSWE v1.1

      DeepSeek V4.1 Flash     74.2
      Claude Opus 5           74.0
      GPT 6 Astra             74.0
      MiMo-V2.6-Pro           71.9
      Claude Fable 5          70.0
      MiMo-V2.6-Flash         67.9
      MiMo-V2.5-Pro           19.0
    • mokre an hour ago

      Maybe you should not trust any of the benchmarks!

    • varispeed 28 minutes ago

      They match my experience. Astra and Fable I rate below Sonnet. They are incredibly poor. They were excellent for a couple of days after release and then plummeted.

      Maybe I am being routed to more quantised versions or less capable models with system prompt to fake Astra or Fable.

  • vatsachak an hour ago

    Wow, the chinese labs are getting good at advertising model releases. The moat is thin.

    Some features of the release I like:

    - Demonstration of diverse tasks, such as using a DAW

    - Graphs from various benchmarks and price ranges

    - Real world use of the model in scientific environments

  • syntaxing an hour ago

    All these new models are such tease for us folks with 128GB of shared memory. Buying another unit now to expand to 256GB is a mortgage payment but it’s getting tempting…

    • verdverm an hour ago
      • trvz 29 minutes ago

        That’s for toy GPUs, like the 5090.

        • verdverm 24 minutes ago

          there are many tasks (increasingly more each day) where small models are more than enough

    • brcmthrowaway an hour ago

      Is there a gamechanger around the corner to reduce DRAM requirements?

      • zozbot234 an hour ago

        You could always stream from SSD storage. Especially effective if you get a cheap old-gen HEDT with lots of PCIe slots to add NVMe storage to and reasonable overall PCIe bandwidth.

        • jkingsman 12 minutes ago

          That nearly certainly boots you to secs-per-tok land (as opposed to tok/s). Plausible if you are willing to wait hours to days for responses for simple testing, but not (debatably) "usable".

      • stymaar an hour ago

        n-gram per-layer embeddings[1][2] might be it.

        [1] https://sebastianraschka.com/llm-architecture-gallery/per-la...

        [2]: See DS 4.1-Flash and Qwen-3.8-Next.

        • verdverm an hour ago

          this is to offload VRAM to DRAM (for GP comment), and makes no difference for URAM

          • zozbot234 an hour ago

            You can definitely offload n-gram embeddings to storage; they're very sparsely used (only a few KB fetched per token) so this is quite effective. Loading to DRAM only becomes necessary if they are a bottleneck to overall performance (which might happen if you're doing very wide batches and everything else uses super fast VRAM/HBM).

            • verdverm 44 minutes ago

              I was looking at the qwen-next-flash, and the weights would fill my OEM Spark on their own, before the n-gram. I'm unclear if offloading to disk can work here, is that what you are implying is possible?!

              • girvo 43 minutes ago

                Check out eugr’s TP=1 sparkrun recipe :)

                It’s an NVFP4 quant, but it fits, and is surprisingly capable.

                • verdverm 41 minutes ago

                  do you have a HF link? HF search is not uncovering it for me

                  (or is it somewhere else)

                  • girvo 17 minutes ago

                    https://github.com/spark-arena/eugr-recipes/blob/main/recipe...

                    This one!

                    I'd recommend pointing your agent at it (after installing sparkrun), and asking it to research the absolute latest in TP=1 Flash-Next - mine grabbed particular vLLM nightlies and mods to improve performance, and it was well worth it.

                    • verdverm 10 minutes ago

                      I have a quirky vLLM on k8s on 2x OEM sparks setup with 9 models available to me. I'm not keen to run nightly vLLM, too many issues with it in the past. Going the qwen-next path means displacing things I use daily :/

                      I have a watchful eye on the diffusion ~ Jev/Kev PR

                      https://github.com/vllm-project/vllm/pull/57250

                      • girvo 2 minutes ago

                        For what it's worth, Flash Next outperforms every other model that is available to us on the GB10 in all of my testing; though if you have two sparks then the TP=2 version is even better and easier (I don't think you'll need the nightly for that at all, just use the recipe)

                        I'm so tempted to buy a second one...

                  • verdverm 19 minutes ago
          • girvo 43 minutes ago

            Nah, I’m streaming ngrams off NVMe on my Spark-alike right now. Works surprisingly well (except for when I accidentally bottlenecked it through my NAS)

            • jkingsman 9 minutes ago

              What kind of throughput do you see on what models?

            • verdverm 42 minutes ago

              interesting, peer comment seems to indicate this is a possibility as well, will have to take a deeper look

          • petu 41 minutes ago

            n-grams can be kept on SSD, no need to hold them in any kind of RAM (at least w/o batching)

          • stymaar an hour ago

            Am I missing a joke? WTF is URAM?

            • verdverm an hour ago

              unified memory, not sure if anyone uses URAM, I human hallucinated it

  • volf_ 29 minutes ago

    I've got a working recipe to run this model on Dual DGX Spark: https://github.com/volfco/spark-vllm-docker/blob/main/recipe...

    Averages ~25-35tok/s which isn't bad for a first attempt.

  • thrownawaysz 41 minutes ago

    >Night 0.8x Usage, 00:00-08:00 -UTC+8

    It's because offpeak electricity is cheaper?

    Funnily it's perfect if you are in the Pacific Time Zone because you can use it daytime 9am to 5pm

  • eriquesito 22 minutes ago

    Funny that all but one video has audio, the house 3D model one, where you can hear (what I assume are) Xiaomi's engineers talking about who knows what.

  • ddxv an hour ago

    This looks great in terms of cost and capabilities, truly pushing the frontier forward in terms of open weight light weight models.

  • MisterMunchkin an hour ago

    I really liked MiMo 2.5, it was really affordable and actually had vision, unlike DeepSeek. (DeepSeek has only recently added it)

    Just tried 2.6 flash on a really niche topic I specialise in and it has done a really good job. They’ve definitely polluted their training data with claudeslop, but looking past the slop there is a decent model.

    • omani an hour ago

      how do you recognize "claudeslop"?

      • Bluestein an hour ago

        It's an honest, load-bearing, simple thing.-

  • DanMcInerney an hour ago

    This is a big week. Probably getting next OpenAI and Anthro models, Grok 4.7, Mimo, etc. These open source model releases are why I can't take the "slow down" crowd seriously. I pitted older Mimo, qwen, step, gpt-oss, and other models against each other playing games like Werewolf and Sketch.io-like games where I let them talk shit while they played against each other. Mimo was by far pareto frontier of game-playing for the models that were <$0.15/m input tokens on OpenRouter. Qwen was pareto frontier in the shit talking game though. Qwen's hilarious. https://www.tiktok.com/@clankerfights/video/7642862917582425...

  • algoth1 an hour ago

    Finally a lab that doesn't cheat on the charts

  • varispeed 30 minutes ago

    These benchmark are useless as they don't say whether they were done before or after Fable and Astra got nerfed.

  • bertili an hour ago

    They mixed up DeepSeek 4.1 Flash with something else on this page, possibly DeepSeek 4.1 Flash means Gemini 3.8 Flash.

  • gigatexal 31 minutes ago

    Leaning into what it cost to train is hilarious and an obvious shot at US frontier labs spending tens to hundreds of millions or more to train their models.

  • alfalfasprout 35 minutes ago

    The moat for OAI and anthropic seems to be very quickly shrinking. Chinese labs are now using RSI-like approaches and even without resorting to heavy distillation they're catching up in a couple of months vs. what would have been 6-12 months a year prior.

    And as these models get better the pace of training is quickly speeding up too.

    This doesn't bode particularly well for anthropic/OAI after they go public.

    • verdverm 28 minutes ago

      token vendors are headed to the same place mobile data vendors went, this is good for everyone but those who thought they could maintain exorbitant prices

  • NooneAtAll3 an hour ago

    does anyone know what unnamed model is on paretto frontier picture right between MiMo 2.5 and 2.6?

    so weird to acknowledge someone being on the front edge, but not name it

  • spwa4 an hour ago

    As for the stats that everyone wants:

    MiMo-V2.6-Flash-310B-A15B roughly GPT-5.6 Luna / Claude 4.9 according to benchmarks MiMo-V2.6-Pro-1.02T-A42B roughly GPT-5.6 Sol / Opus 5 according to benchmarks

    Perhaps with IQ2 flash will run on 128G M5?

  • omani an hour ago

    ah, would you look at that. I was wondering why mimo 2.5 became "dumber" the last weeks. I was speculating they are probably about to release a new version of the model. because the model really acted out a lot. especially the last two weeks. dont know, was just a feeling, highly speculative.

    but now I got my "proof".

    • sandblast an hour ago

      I guess that would only be possible if your provider was Xiaomi itself?

      • omani an hour ago

        yes. I use opencode and opencode uses Xiaomi as a provider.

  • jwpapi an hour ago

    In the chart they use "Pareto Line", which I think is wrong. Pareto is 20% effort leading to 80% results. Which could be interpreted as models costing 20% having 80% of peak intelligence, but that’s not what it looks like to me.

    It looks like the "Frontier Line" to me, which is also often misinterpreted. frontier does not mean the best models. It means all models that are not strictly dominated, meaning in most cases: Not same price or cheaper and more intelligent.

    I personally would like the word frontier to be used with more criterias: Open Weights, per use-case, etc etc. This would make model selection easier, but I understand it’s not an easy thing to do.