Hacker News Reader

Comment by elfbargpt | Hacker News Reader

I've always been surprised Kimi doesn't get more attention than it does. It's always stood out to me in terms of creativity, quality... has been my favorite model for awhile (but I'm far from an authority)

View on HN

It’s good, but it’s not quite Claude level. And their API has constant capacity issues.

Price/quality is absolutely bonkers though. I loaded $40 a few weeks/months ago and I haven’t even gone through half of it.

It has long been Claude level since 2.5

Why use China model API from China if there are many independent providers available via Openrouter?

Openrouter will route to china hosted models when there are US hosted providers of the same model. Is there a setting to set your preference or to blacklist providers like alibaba cloud for example?

I use OpenCode and the openrouter provider. From opencode I only select the model like kimi-2.6 and have no way of selecting which cloud hosting will receive my request.

Settings > Guardrails > [your workspace] > Providers + Block provider

Yes, you can blacklist providers in OpenRouter account settings.

Yes, you can globally ban providers in your openrouter settings.

to support the companies that open source their models

It's also one of the few models that seem capable of drawing an SVG clock

https://clocks.brianmoore.com/

Interesting that the best performers are all Chinese-made models (DeepSeek and Qwen also perform consistently well). I wonder if there's more focus on vision and illustration in their training, or if something else is leading to their clear lead on this one test.

Is it? In your link it definitely failed to draw the clock.

It redraws it every minute, and some models give quite different results although the prompt is exactly the same.

This reads like satire, but I've been feeling that a lot lately.

I'm not really sure how this works, but I stayed on the page for a while, and then it reloaded and all clocks changed. I guess there's either a collection of different clocks generated by models, or maybe they're somehow generated in the real time, but the fact is what you see is not necessarily what I see.

It reruns a prompt every minute to all the models included. Everyone is gonna see something different but I've spent too long on it and there's a consistent pattern of Qwen and Kimi outperforming the others

This site was made months ago and it seems its only been updated with the latest model of a couple of the providers so keep in mind that many of the Chinese models haven't been updated

Seems like it regenerates them to reflect the current time. Funny to see how some models (like Kimi and Deepseek) sometimes get it right and other times fail miserably on the level of ancient models like GPT 3.5.

It reruns the prompt every minute.

Dirt cheap on openrouter for how good it is, too. Really hoping that 2.6 carries on that tradition.

Kagi has it as an option in its Assistant thing, where there is naturally a lot of searching and summarizing results. I've liked its output there and in general when asked for prose that isn't in the list/Markdown-heavy "LLM style." It's hard to do a confident comparison, but it's seemed bold in arranging the output to flow well, even when that took surgery on the original doc(s). Sometimes the surgery's needed e.g. to connect related ideas the inputs treated as separate, or to ensure it really replies to the request instead of just dumping info that's somehow related to it.

I remember when the first K2 dropped

It was the best creative writer by some distance

Maybe because it's a bit of like unleashing a chaos monkey on your codebase? I tried it locally (K2.5 72B) and couldn't get anything useful.

Huh, that's not a thing?

The parent poster is probably referring to Kimi-Dev-72B¹, which is a much smaller and older model, while people are probably more familiar with the big and fairly powerful 1100B Kimi-K2.5².

[1] https://huggingface.co/moonshotai/Kimi-Dev-72B

[2] https://huggingface.co/moonshotai/Kimi-K2.5

Yes it was good for its time, but 10 months old now which is a long time ago in this space. It was also a fine-tune (albeit a good one) of Qwen-2.5 72B.

I wish they did more smaller models. Kimi Linear doesn't really count, it was more of a proof of concept thing.

Reply on news.ycombinator.com