Item — HN Reader

Xiaomi have been cooking a lot in recent times. Their model, especially the pro series, is underrated in my opinion. It haven't received the attention it deserves while it is pushing higher and higher in benchmark scores (looking at artifical analysis), and this was before Deepseek dropped V4.

Furthermore, their pricing plan is insanely cheap, they even upped usage limit for their cheapest plan, lite plan, which is at 5$ / month. And now, they are dropping a Harness for their own model? Amazing. I wish they added support for installation through Homebrew though.

On another note, this is what I would like to see more of from a company, what I do not welcome is startups making their model exclusive and hurt their customer base through sabotaging as a way to prevent eventual distillation attempts.

2 comments

>Furthermore, their pricing plan is insanely cheap, they even upped usage limit for their cheapest plan, lite plan, which is at 5$ / month.

Unless something changed their plans aren't really worth getting. They're not that much cheaper than the per-token rates, and because it's a plan, you have to contend with weird usage restrictions. You're better off paying per-token unless you have some use case that demands a very steady stream of tokens.

Wow. I saw 4.1B credits and thought it was super generous. But my math says the subscription plan gives less value than the API.

For example, API input is $0.435 / M tokens, which works out to 13.79 M tokens for $6.

Plan is 300 credits per input token, which works out to 13.67 M tokens at 4.1B credits per $6.

Very similar math for cache input and output.

Indeed. I did the math and arrived at the same conclusion. They don’t really subsidize their token plans. Maybe because their api pricing is already dirt cheap

Looks like they have very effective collaboration with DeepSeek and Kimi. Those three models have been bouncing ideas and sharing R&D innovation, which made all of them improve very fast.

Based solely on quality and price, OpenAI, Anthropic, and other western models just can't compete with the new generation of Chinese open models.

>Looks like they have very effective collaboration with DeepSeek and Kimi.

The collaboration is informal. People don’t seem to realize this, but the Chinese internet for programmers and developers today feels a lot like StackExchange in its heyday. There’s a huge emphasis on sharing knowledge, because sharing what you know builds your profile, and becoming a rockstar in a subfield is one of the only ways to get ahead.

Competition in China is ruthless. But unlike in North America, where individuals are often bound by agreement to hoard knowledge because it can give them a competitive edge, the competitive advantage in China is building face and peer recognition. And that comes from proving that you are worthy of being a "master/teacher", and that extends to the valuation of your knowledge business. For example, the third wave coffee shops in China, the master roaster is often called "master/teacher" once they win a roasting competition and start sharing new knowledge of roasting in the public sphere, and that's a title of sincere respect.

You can see parallels with those that apply to give talks at conferences and post snazzy technical presentations they give in the US, but the bar for what qualifies as new knowledge is far higher in China because there's a massive ecosystem of people rushing to outcompete what you have to offer, and once the ball gets rolling on knowledge sharing, lots of people will go off and build upon that knowledge or try to build businesses on top of that, which in turn produces more knowledge.

Reading developer forums in China, once you crack the code (I find Gemini will get you a good chunk of the way with good translations), they are really quite far ahead with what they're willing to share. And I suspect in great part, the decision to release open-weights is heavily tied to that concept of building face/peer recognition = building valuations.

Thank you for the insights, this sounds like an environment I want to partake.

> Reading developer forums in China, once you crack the code (I find Gemini will get you a good chunk of the way with good translations), they are really quite far ahead with what they're willing to share.

Are you able to share those forums and other resources? I would love to read what people in these communities are sharing.

What forums do Chinese programmers use that you would recommend exploring around in?

Mainly want to see the people who are building stuff I've never even thought of.

> huge emphasis on sharing knowledge

> developer forums in China

Care to share a few such forums or websites, genuinely curious as someone who reads basic Chinese.

A decade ago, I knew there were CSDN, jvjin (may not be the correct pinyin) where people talked about tech. Nowadays they have just become SEO farms like the rest of the internet.

Actually there are very few Chinese forums left, because it is a huge hustle to run one, and the same rule applies to the comment section of any personal blogs.

Right now pretty much the only one left with a bit discussion is v2ex, but it's blocked by gfw and kinda becoming a ranting corner for programmers. Some also recommend nodeseek and a couple other new sites, where you can find shady stuff but not tech stuff. Then you have tech blogs of big corps, or personal ones, but still not many quality tech posts. There are also some on WeChat but it's kinda cumbersome to use so I stopped.

> Looks like they have very effective collaboration with DeepSeek and Kimi. Those three models have been bouncing ideas and sharing R&D innovation, which made all of them improve very fast.

Very fascinating to learn this, didn't know Moonshoot (Kimi) also collaborated with others. I think I read in another post that DeepSeek and Qwen team shared the same building? So that kind of explains it.

> Based solely on quality and price, OpenAI, Anthropic, and other western models just can't compete with the new generation of Chinese open models.

I have to agree. I had the great opportunity to take the offer Z.ai had with their Christmas deal, their lite plan was 3 months for 7$. GLM-4.7 was already impressive enough.

When they released GLM-5-Turbo and GLM-5.1, that is when I came to the realization of how close the gap is between proprietary western models and Chinese open-weight ones (not all of them are ofc).

I could barely believe how good GLM-5.1 was, I didn't think I was using it in CC and had to check the settings again. It's astonishing how close the gap is now, and this competition benefits us very much, the pricing is so low atm, its amazing.

Consider also Chiness EV prices. Unfortunately you can't buy them in the US it seems. https://www.reuters.com/business/autos-transportation/averag...