Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per second

mimo.xiaomi.com

Post reply on HN
Print view View original

341–350 of 512 posts

Re: MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per second

#341

Post by wild_egg »

andai wrote: post #340
throwaway67678 wrote: post #170

Earlier quoted context omitted.

Agent mania setting in It's also pretty funny sometimes how it gives weird future roadmap estimates ("part 2 - 3 weeks, part 3 - 2 months", etc.) and when you tell it to actually do those changes it's pretty much done in half an hour

I heard an anecdote. Guy spent several days trying to convince his AI agent to build a feature. Kept saying it was crazy complicated, would take weeks. Finally he convinced it to try. It one shotted it in 30 seconds. Turns out the agents' idea of what is hard and easy also comes from Common Crawl.

Why on earth would you spend any time at all convincing an agent of anything? You say "just do it" and off it goes.

Re: MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per second

#342

Post by InterviewFrog »

Terretta wrote: post #248
smith7018 wrote: post #184

Earlier quoted context omitted.

I've long believed those numbers were faked by Anthropic/OpenAI to serve as a form of advertisement. The estimates are impossible to verify and their ability to do "2 days of work" in 10 minutes will presumably make the user go "Wow, I just saved SO much time!" Plus, the unnecessary text eats up the users' tokens so it helps the companies on the backend, as well.

> the estimates It doesn't estimate. It generates tokens that read like estimates associated with the context in its training material. What would you expect the generator to output instead?

This is so 2023. The thought process.

At that time the predominant view was that LLMs were nothing but stochastic parrots, that they would plateau, and that hallucinations couldn't be fixed.

At this point I doubt there are any AI sceptics left. That ship has long sailed. The only thing that matters is whether the estimates are accurate, and AI can improve on that too.

Even humans only estimate based on neurons firing in prior patterns.

Re: MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per second

#343

Post by wild_egg »

irthomasthomas wrote: post #287
ghshephard wrote: post #271

Earlier quoted context omitted.

I think people are continuing to view these systems as pure LLMs - when that ship sailed 6+ months ago. Between being able to review memory, using agent harnesses and sub agents and skills to go out and discover information - modern systems (Codex, Claude Code, Cursor) - use LLMs - but the LLM is only a small component of it. Compare what you get from sending a request to a chatbot like ChatGPT - to what you can from…

No one is bitter lesson pilled anymore. Everyone is pivoting to neurosymbolic systems. It looks like Gary Marcus was right.

How is neurosymbolic not aligned with the bitter lesson? The bitter lesson is completely agnostic to architecture.

Re: MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per second

#344

Post by digdugdirk »

gertlabs wrote: post #282
unrvl22 wrote: post #191

Earlier quoted context omitted.

why is deepseek v4 pro a lot lower than flash? where is mimo 2.5?

DeepSeek v4 Pro struggles with a custom harness, and all the models ranked above it don't, so it gets downweighted in the agentic coding benchmarks (although it ranks better than Flash in one-shot problem solving: https://gertlabs.com/rankings?ow=1&mode=oneshot_coding ). We ran plenty of samples. MiMo v2.5 is on there, as well as the pro version. We found a few anomalies in our evaluations, which makes sense -- if ev…

Can you explain more about how it struggles? I haven't noticed any issues in my usage, so I'm just curious what is meant by this.

Re: MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per second

#345

Post by dr_dshiv »

wild_egg wrote: post #341
andai wrote: post #340

Earlier quoted context omitted.

I heard an anecdote. Guy spent several days trying to convince his AI agent to build a feature. Kept saying it was crazy complicated, would take weeks. Finally he convinced it to try. It one shotted it in 30 seconds. Turns out the agents' idea of what is hard and easy also comes from Common Crawl.

Why on earth would you spend any time at all convincing an agent of anything? You say "just do it" and off it goes.

Ya, but “doit” is 2x more efficient

Re: MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per second

#346

Post by nl »

throwaway894345 wrote: post #39
amunozo wrote: post #17

These price and speed optimization from Chinese providers, combined with the raising prices from American ones will change the game sooner than later. Many companies are finding issues with the AI bills already.

I wonder what are the economics driving these pricing decisions? Are the Chinese companies just subsidizing their models to a greater degree than the US, or is this an emergent property of energy policy between countries?

Their models are much smaller: 1T vs 5T for the frontier models. 1T is Sonnet/Google Flash size, not Opus size.

The $0.87/M tokens price for Mimo Pro is probably subsidized.

Mimo models aren't widely available on western providers, but Kimi and Deepseek are similar sizes and cost about the same to run. They are priced $3-$4/M tokens (which is right were Google's very confused range of Flash models are priced at: between $0.40/M tokens and $9/M tokens depending on exactly which model - and you don't want the $9 one!).

Anthropic overprices Sonnet (probably because of their capacity issues). GPT 5.4 mini is $4.50/M tokens.

https://docs.fireworks.ai/serverless/pricing

https://www.together.ai/pricing

Re: MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per second

#347

Post by acdha »

solenoid0937 wrote: post #70
alkyon wrote: post #44

Earlier quoted context omitted.

Sounds like exponential growth of crappy software. I'm not saying that before we didn't have mass produced crap in SE, but now it will turn into explosive overflow.

Crap is fine if it gets the job done. I think software as an industry will change to more ephemeral construction.

What counts as “done” has a time component, so I think we’re going to see more of a spectrum where some businesses try to skimp as much as their market will allow but others will recognize that racking up technical debt is a long-term loss. Stuff like brochure sites will certainly be cut down but anything where there’s liability or long-term customer relationship is going to need to factor in quality as well.

Re: MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per second

#348

Post by seewhydee »

rubyn00bie wrote: post #289
behnamoh wrote: post #213

Earlier quoted context omitted.

Same. How can DeepSeek serve the V4-Pro at such high speeds despite the sanction?

The sanctions only “prevent” them from directly buying NVidia’s latest and greatest in the sense that NVidia can’t sell directly to them. Essentially, there are companies now who are in a country without the sanctions, they buy from NVidia (or a partner), and then ship them off to China. For the orgs in China doing this, there’s zero legal risk besides having foreign customs service intercept the shipment and losing…

That wouldn't explain why Deepseek is fast relative to other Chinese providers, especially considering that they're reportedly ahead of the curve among Chinese companies in moving off Nvidia. I think their quant fund background has more to do with it. Their models are clearly designed with performant inference clearly in mind.

Re: MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per second

#349

Post by pmontra »

dakiol wrote: post #136

So, regarding the productivity argument: I don't get it. It doesn't really matter (for regular employees) that you can do now in 2h what before it took 2 days. Why? Because it's not that you have the rest of the day for yourself. You still have to work 8h/day as usual. But now the pattern is different: instead of enjoying the craft digging deeper into problems in the span of 2 days, now you are rushing into some slot…

If you split the tasks for the AI in small chucks you keep the architectural control and it's not a slot machine anymore. You still read code and occasionally you write code too. Not much but it's the price to pay for the extra speed.

If you start the AI on something big and come back after one hour then yes, you might discover that you wasted an hour and got nothing.

Re: MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per second

#350

Post by pmontra »

dd8601fn wrote: post #296
linsomniac wrote: post #217

Earlier quoted context omitted.

Odd, I'm having the opposite experience. The thing I really love about working with computers is when I achieve something. That's the thing that makes me figuratively, and sometimes literally, throw my fists into the air and go "Yeaaah!" With the AI tooling, I'm getting those more like a couple times a week. Plus, I'm using AI to attack the things in my day that are "a drag", and getting them done too. The highs are…

I did a deep binge on two or three projects I would never do, and like five small ones that would have consumed months. It felt like that, kinda, for a bit. Now whenever it does something for me I get nothing. I didn’t do it… the chatbot did. What’s for me to celebrate? How can there be any real pride or satisfaction for a thing that was just handed to me because I asked for it? If anything it diminishes my satisfact…

> The things I had to learn and the informed decisions I had to make? All pointless trivia, now. A child could do it.

Probably this is a hyperbole. Did you do the experiment? I expect that the child won't be able to do it. Ask an adult. Same thing. Ask an expert of the domain. Maybe but not as fast or as good as you.

Post reply on HN