Inference Optimization for MiMo v2.5: Pushing Hybrid SWA Efficiency to the Limit

https://mimo.xiaomi.com/blog/mimo-v2-5-inference

116 49 16 days ago

mimo.xiaomi.com

LaurensBER2• 12 days ago

It's really cool and interesting to see the kind of engineering that goes into Xiaomi (and Deepseeks) inference optimizations. Z.ai has also published some interesting papers although I haven't had a chance to go through them yet.

It does inspire hope that the Chinese labs seem to be so open although the sceptic in me does wonder what their end game is.

Surely, from a purely economic perspective it would be wiser to keep this proprietary and benefit from the increased API traffic?

ignoramous• 12 days ago

the kind of engineering that goes into Xiaomi (and Deepseeks) inference optimizations

At Xiaomi, MiMo is now led by Luo Fuli. She is a former Alibaba & DeepSeek employee: https://newsen.pku.edu.cn/news_events/news/people/15385.html (https://archive.vn/I8Pmu) / https://e.vnexpress.net/news/tech/personalities/who-is-luo-f... (https://archive.vn/sb3B6)

Don't know if it is due to Luo, but it is striking how similar performance & pricing of the models, DeepSeek v4 Pro & MiMo v2.5 Pro, is.

xyzsparetimexyz• 12 days ago

Christ, she had such a nicer vibe than Altman or Amodel.

bee_rider• 12 days ago

The bet could be that they’ll ultimately be able to sell hardware capable enough of running local models comfortably.

notnullorvoid• 11 days ago

If China is able to undercut Nvidia on high performance local AI hardware, they will pop the AI bubble in a matter of days.

They wouldn't even need to make something equivalent to the latest hardware. A Chinese RTX 3090 equivalent would be enough.

onlyrealcuzzo• 12 days ago

Standard commoditize your complement.

trollbridge5• 12 days ago

Their game? Sell me tokens instead of me buying them from an American lab for a higher price.

Publishing open weights gives me more confidence in the model, and ironically makes me less anxious about making sure I can replace the cloud usage with a local alternative. Whereas I’m very nervous right now with relying on 5.6-Sol - what if they triple the price, nerf it, etc.?

TurdF3rguson5• 12 days ago

Publishing open weights gives me more confidence in the model

Why? It's not like you can audit weights like you can with code.

what if they triple the price, nerf it, etc.?

What if an open weights infra provider does that? What's the difference?

yogthos2• 12 days ago

What Chinese firms are doing makes perfect sense from the commercial perspective actually because they understand how a classic commoditization spiral works. The reality is that models themselves are general commodities and there's just not enough difference between them. A company can get ahead of others by a few months, but then the rest quickly close the gap. It's a really low margin business because there's no way to differentiate yourself.

Chinese companies know that there's no profit in general purpose models in the long run, and they're treating models as shared infrastructure akin to Linux. They're amortizing the cost of research by keeping models open, and rapidly closing the gap and driving prices towards the marginal cost of inference. The money is going to be in customization niches. Companies will charge to tune models for specific use cases and charge support for that. There's also going to be money at the bottom for hardware vendors making chips and memory. But the middle tier of generic LLMs is seeing involution where there's relentless competition driving profits towards the bottom.

CuriouslyC3• 12 days ago

The Chinese labs incentives is to run inference for the world, because inference can run on the homegrown Chinese chips (giving them a guaranteed market for their hardware) and they have cheap plentiful power.

The US frontier labs have an incentive to do deals with large firms to act like a contract research organization, taking royalties on creations/discoveries. Alex Karp called this out in his rant ("Why charge for tokens, take a %") and he's basically right about this.

alightsoul4• 12 days ago

its a governement mandate that states that AI research must be open source ,that's one benefit of communism

yorwba2• 11 days ago

There is no such mandate. ByteDance keeps their models closed. So does iFlyTek. Qwen Max is closed as well.

ricardobeat2• 12 days ago

Such a well written article, refreshing to read in between all the slop.

I've used Mimo extensively in the past few months, can't wait to see what v3 will bring.

trollbridge5• 12 days ago

Are you using UltraSpeed? It's my favourite thing about MiMo.

ricardobeat2• 11 days ago

I did, it’s amazing. Shame it’s going away and not available permanently.

zozbot234• 11 days ago

It's a very nice summary of KV cache storage requirements for the leading open weights models, but ultimately it still shows the DeepSeek V4 series winning by a huge margin on that front. And DeepSeek actually does quite well on needle-in-haystack context recall benchmark so it seems like the outstanding context compression it has isn't even impacting smarts all that much. Keep in mind that KV cache memory requirements usually has a direct impact on how much you can parallelize inference on both large-scale platforms and lower-end edge compute, so this is a very important metric.

killingtime74• 12 days ago

Will they lower the price or is this documenting past work

limbicsystem• 12 days ago

Their pricing is incredible on the token plan - something like 50b tokens for $60!

polski-g• 12 days ago

No that's not right. It's 50b credits, not tokens. What is a credit? Nobody knows.

Someone did the math a few months ago and paying API prices was the same as the monthly subscription.

amanharshx• 12 days ago

This is really neat. They've done some really impressive engineering here ngl with the ~95% KV cache hit rates. MiMo and Deepseek both do seem to get the job done for me. Hope they can keep this pace while staying open source.

SwellJoe2• 12 days ago

I've been saying efficiency is the next "frontier" in AI, at least for LLMs that people use daily. Companies have started to really balk at token costs from the major providers, and there's some evidence that the cheaper Chinese models are chipping away at Anthropic/OpenAI dominance from below (as cheaper Chinese products have done in other industries for many years). I continue to think that the vendor that figures out how to efficiently serve a Good Enough model at a much lower price will be the one to win in the end.

DeepSeek is remarkably efficient at caching and their cached token rates are crazy cheap; using it with Reasonix is free real estate, like 97% cached tokens, ends up costing like 30 cents an hour to use DeepSeek V4 Pro. I hadn't dug into MiMo's caching behavior as I haven't used it as heavily as DeepSeek, but this indicates it's close to DeepSeek.

At this point I don't see a reason to use Sonnet, Haiku, or the smaller GPT models, because their API rates are much higher than the best models from MiMo and DeepSeek.

We're still figuring out the upper bounds of capability and I am still finding next generation models are unlocking things I couldn't readily accomplish before and I'm willing to pay more for them (at least, I'll pay the $100 or $200 subscription rates for them, I couldn't justify the token expense for most of my dev work), but we're already at a point where someone building standard CRUD web apps doesn't need the top models and probably doesn't benefit much from using them.

mapontosevenths2• 12 days ago

I agree, but maybe for different reasons. I think Karpathy is right. We need models that reason, not models that memorize.

Karpathy calls it a "Cognitive Core", and it's essentially a small model that learns to reason and look up the data it needs as opposed to a giant model that memorizes all the data in the world and tries to process large chunks of it all at once with every thought. I think it will be based on the thing that grokking, the lottery ticket hypothesis, and the universal weight subpspace hypothesis all point to.

Eventually someone will figure out how to build it and the entire economy that we've now built on top of the wacky idea that nothing can possibly ever get more efficient will collapse overnight.

Sometimes I wonder how much Nvidia would pay someone not to release a thing like that, and then I wonder if that's already happened.

CuriouslyC3• 12 days ago

The theorem you want to pay attention to is the no free lunch theorem. The important thing to understand there is that the larger models give you "free lunch" in the sense that you can approximate more different systems accurately at the cost of model size. If there was a Karpathy style universal solver, it wouldn't be very smart unless we scaled it up.

This isn't to say that there aren't a fair amount of wasted parameters in current LLMs, but then we already kinda knew that since you can quantize models down to 3-4 bits per weight often times with minimal loss.

SwellJoe2• 11 days ago

I don't think LLMs get us there, at least based on my understanding of them. You don't get the reasoning at a very high level without a similarly large parameter size.

Though, I don't think we're really at the maximum density of intelligence per byte. Small models of today are approaching the frontier of maybe 2-3 years ago. Gemma 4 and Qwen 3.6 in their dense versions feel quite capable, give them a search tool and they'll research a topic and often give a pretty good explanation of it. They understand more and hallucinate less than GPT-3, GPT-3.5, and maybe even GPT-4 (the first probable trillion parameter model), and their tool capabilities opened the door for agentic use.

But, those small models are still smaller thinkers. It's not a cognitive core that can do anything if given a way to look up the data, though they're more capable than a lot of folks realize.

Also, there's no reliable way to look up data, anyway. The web is poison, approaching the point where naive web search will most likely return misinformation or intentional disinformation. The source of all human knowledge must be viewed as adversarial, now, which makes having thinky machines that search the web and do stuff with the results a risky business.

nmfisher2• 12 days ago

Related, I was given access to mimo-v2.5-ultraspeed, which is amazing. This is now my expectation for speed, it’s fast enough for me to stay mentally engaged rather than getting distracted waiting for the agent to churn.

embedding-shape• 12 days ago

The -spark variant of GPT was a ton of fun indeed, such a shame it's so dumb though so really hard to rely on. If you were to compare the quality of mimo-v2.5-ultraspeed with anything from OpenAI/Anthropic, where would it be placed ~more or less in your view?

perching_aix• 11 days ago

Is it the same quality as base mimo v2.5, or different? I've been enjoying regular mimo v2.5 quite a bit via opencode, if ultraspeed provides the same quality, that's crazy.

nmfisher2• 11 days ago

I'm not a regular v2.5 user, so I don't really know. But given the TileRT team write-up says the entire network gets quantized to FP8 (and experts get quantized to FP4)[0], I'm assuming there's at least a modest drop in quality.

[0] https://www.tilert.ai/blog/breaking-1000-tps.html

Inference Optimization for MiMo v2.5: Pushing Hybrid SWA Efficiency to the Limit

Related Stories