[Tiberium]

More details:

- https://platform.kimi.ai/docs/guide/kimi-k3-quickstart

- https://platform.kimi.ai/docs/pricing/chat-k3

1M context, pricing is $3/$15 for 1M tokens (cache $0.3), which is extremely high for a Chinese open-weight model, but if it's truly competitive with most of the current frontier and is only behind Fable/Sol, the pricing is justified.

This is 1:1 pricing of Anthropic's Sonnet series (except Sonnet 5 which is currently on discount), and very close to 5.6 Terra pricing (Terra's input is $2.5).

One thing to consider, though: reasoning efficiency matters directly for how expensive a model actually is in real use. GPT's models are extremely reasoning efficient, and some Claude models like Fable at lower effort are as well. So if Sol spends 10K reasoning tokens to do something (at $30/1M) vs Kimi K3 that spends 50K reasoning tokens, Sol would win on cost effectiveness.

[dghlsakjg]

Tokenizers also matter. Anthropics tokenizers will encode the same piece of text at a way higher token count than OpenAi, for example.

That said, Kimi is competing against GLM in my mind, and GLM 5.2 is less than 1/3 the price.

[mdasen]

It also depends on how many tokens it needs to burn through to accomplish something.

At this point, I always look at things like Artificial Analysis' total cost to run their tests. It'll take into consideration the cost of tokens, how many tokens it burns through, and how effectively it uses caching (and the price of that caching).

If a model "costs the same" but its reasoning ends up going through a ton more tokens, it doesn't really cost the same in real world usage.

[KennyBlanken]

Precisely. GLM 5.2 Thinking is pretty damn good - but it regularly does something nonsensical. Or even spits out what looks like a fragment of its memory cache. Or returns a bunch of Chinese.

I find myself having to resubmit a query very often...so it being a third of the cost of other AIs isn't really relevant.

[cmrdporcupine]

GLM is actually quite expensive in actual practice because it's not very token efficient. I've yet to find a way to run it on a monthly sub reliably for cheaper than Codex.

Neuralwatt was cheap (but slow) but they cranked their price.

Ollama monthly sub is speedy but doesn't offer a lot of quota.

Right now unless you're paying by the token, there's no cost based reason to use the open weight models for daily coding work because the monthly coding plans from Anthropic and OpenAI are a better deal.

[computerex]

I know GLM is relatively expensive and so is Kimi, in comparison to those DeepSeek V4 pro and flash are a godsend and are absolutely good value.

[arizen]

And DeepSeek V4 Flash + GLM 5.2 is a really good blend of both (fast/cheap DS + more intelligent GLM)

[versteegen]

I think MiMo 2.5 seems to be better than DS4 Flash, at the same price. DS4F can write pretty advanced code but it way overthinks the simple stuff, its CoT is full of errors (immediately corrects itself), and yesterday I was shocked it reliably struggled to spot a misplaced % in a format string when reading a whole file.

[calgoo]

Exactly, I design in glm 5.2, and then build with deepseek pro. I pay deepseek by token count, and honestly I added $20 and it just keeps going, hardly burning anything.

[pimeys]

I use V4 flash as my personal agent. It categorizes documents, organizes my calendar, searches information etc. for pennies. Amazing model.

Not very good for programming though.

[UncleOxidant]

I'm on the Z.ai quarterly subscription plan (got in when the price was lower) and I was using it through opencode and it was like I'd only get maybe an hour of usage (if that, sometimes) before it would time out and say come back in 5 hours. Now I'm using it through their Zcode harness and I rarely hit that - they say they're giving 1.5x usage if you use it through Zcode, sometimes seems like even more than that.

[esafak]

Too bad Zcode sucks. The promo ends this month anyway.

[KronisLV]

> I've yet to find a way to run it on a monthly sub reliably for cheaper than Codex.

Matches my experience, I got their Pro subscription and while I enjoyed the model itself a lot and while their ZCode harness is also pretty nice, it gave me less tokens for similar amounts of money that Anthropic would give me on a subscription: https://blog.kronis.dev/blog/z-ai-s-glm-5-2-is-a-great-model...

I'm yet to try out Kimi, but if their subscription were to be anywhere comparable to Anthropic/OpenAI, I might just switch over because competition is good.

DeepSeek V4 Pro is really affordable per-token but regularly kept making mistakes in the tasks I gave it. I mean I could at least afford the tokens to go over the work a 2nd, 3rd, 4th and 5th time and gradually fix most of the issues, but it was a very frustrating mode of work.

[stavros]

I don't know, DeepseekV4 is so dirt cheap that it makes lots of sense to use over Sonnet.

[mark_l_watson]

re:

> Right now unless you're paying by the token, there's no cost based reason to use the open weight models for daily coding work because the monthly coding plans from Anthropic and OpenAI are a better deal.

Maybe. I am on a $20/month Anthropic subscription this month but I also use Claude Code frequently with Deepseek v4 flash and pro, GML5.2. For simple work Deepseek v4 flash is so nice because it is fast.

What you say is true however, the US hyper-scalers are still (desperately?) subsidizing subscriptions for market share to boost there valuations.

I really want to see AI inference costs approach zero, and I think I just need to wait a few years to see that.

[cmrdporcupine]

DeepSeek is a whole other story. It and a few others are quite economical. But they're also not nearly at the same level.

I can get by working on code strictly in GLM. I can't with DeepSeek. It makes some pretty careless mistakes and isn't a very deep thinker.

It is very useful as a general purpose model for non-coding purposes though.

[humanfromearth9]

For maths, it's also wrong most of the time. Generated stuff looks right, then let Opus audit and see the disaster.

DS4 is usable iff you have a way to test the generated stuff, and to convince yourself that its production is right. With successive review-fix rounds, it's obviously way more reliable too, but that can't compensate for it's lack of rigor when reasoning. It's very smart but neither rigorous nor careful. And that is a direct result of its architecture.

[ifwinterco]

I found this with kimi k2.7 as well: on paper it should be quite cheap, but it's not because it uses a lot of tokens for quite simple tasks

[tokai]

Compared to the flagship models GLM is still a 1/10th the price on the task I have tested.

[cmrdporcupine]

Only if you're paying for said flagship models through API prices. Which I specifically said I was not and specifically mentioned coding plans.

[InsideOutSanta]

> That said, Kimi is competing against GLM in my mind, and GLM 5.2 is less than 1/3 the price.

Having used GLM 5.2 extensively and K3 for a few hours now, these models are nowhere near each other. 5.2 is a great model, and I use it for a lot of things, but it's noticeably below Opus 4.8 or GPT-5.5 in real-world usage.

K3 is in the same ballpark as Fable or Sol.

[leecommamichael]

Tokenizers define the alphabet on which the language model is trained. I don't want people to get the impression it's a module which can be swapped out or modified on its own. Alphabet size is a design consideration related to correctly encoding the training data.

[smallerize]

That's true, but it makes it difficult to compare pricing when it's based on tokens. Maybe we need a benchmark for price per a specific input, like enwiki8.

[leecommamichael]

Yes, almost all work people share which seeks to measure the capabilities and differences of models needs to get more precise. We are clamoring to say something meaningful about these things.

[victorbjorklund]

It is kind of a shame we ended up comparing token pricing across models and providers when it doesn’t really make sense. Not sure what would be better though.

[alain94040]

Use price per page (standard English text)? That would also help make the metric easier to visualize.

If you think a page is too vague, use a famous known writer's work as a reference.

[whoopdeepoo]

Well isn't that what benchmarks are for? They compare total cost for a unit of work.

[whodatbo1]

A better metric is price per byte. Most thinking traces, prompts, skills are in plain English, which is roughly 1 byte per character, assuming UTF-8 encoding (even code should not be much more either). As an aside, it is common to use bits-per-byte as a loss metric instead of the per token calculation, precisely because of the effect of different tokenizers.

[smallerize]

It's going to vary dramatically based on which text you put in. Really it's hard to make one benchmark number that's relevant to all cases. But maybe we can make something a little more specific, like regular English text, code, the model's own thinking tokens, image inputs etc.

[kevincox]

But even that isn't the whole story because the models can produce wildly amount of thinking output as well as regular output for a similar query. Sometimes you can take a cheap model and have it think a ton or an expensive model that thinks little and get similar results. But the number of tokens generated will be wildly different.

[semiquaver]

I’ve been struggling to understand the reason for the newer apparently less efficient Anthropic token encoding. If all inputs are less efficient in this encoding, why does it exist? Has Anthropic released any information that would convincingly show it was anything other than a stealth price hike? Please don’t respond if you are speculating.

[remus]

> Please don’t respond if you are speculating.

I doubt you are going to get a response from an anthropic employee, but I think it is safe to assume they have swapped to a new tokenizer because it improves the performance of their models.

[re-thc]

> the reason for the newer apparently less efficient Anthropic token encoding

Less efficient in token usage but per the blogs; it enables the model to perform better.

[mnicky]

It's really simple I think.

More tokens per same text length means more capacity to encode information. More information means model can potentially perform better.

They introduced it around the time the Mythos came so my speculation is that if you have more capable model at some level you may find the current information encoding not using its full potential.

We will see whether OpenAI also introduces new tokenizer when they come to Mythos-size models.

[asenna]

With that kind of pricing, I don't think they're competing with GLM with this new launch.

[wmedrano]

Where does the less than a third the price come from? From the provided benchmark suite, it came slightly under half.

[zvikara]

I believe Kimi is spending more on marketing than GLM (a lot of ads lately) so I guess that's part of what the higher price supposed to cover.

[ImageXav]

I've been avidly using Fable since it was re-released and while it has been excellent at building the apps I want, the reasoning has been completely opaque.

Kim, however, has exposed the whole reasoning trace, or enough of it to matter. I'd almost forgotten how nice it is to see this. I've been able to see all of the weird twist and turns it takes and it is joyful. But also, far, far more informative and means I can debug ideas far more thoroughly. Also, at a first glance it seems to have gotten quite far on a niche hobby horse of mine that no LLM has been able to crack. I'll be testing this more for sure.

[epistasis]

I have severe complaints about Anthropic's product managers on this front. Their preference for hiding, obscuring, and trying to wrest control from the user are a bit harrowing. It would be wonderful to go back to Claude Code from before March. It seems like every release destroys value for me!

[qeternity]

It's a defensive tactic to reduce the effectiveness of distillation.

Say of that what you will, but it's not because they want to wrest control from users.

It's because they don't want Chinese companies to do exactly what Moonshot (Kimi creators) and others have done.

[anon373839]

Anthropic’s position being that it is entitled to train models on the creative works of anyone at any time, but its own slop generators’ outputs are sacred jewels that must be protected from being learned from.

[program_whiz]

Its not surprising a thief would be the most paranoid in securing their spoils.

[desterothx]

According to Pat Toulme, the thinking traces and outputs that the Chinese researchers distilled are useful for getting initial trajectories, to prevent a cold start during RL. Once you get those initial correct trajectories (the model actually solving a task), you can generate you own new traces, so the distillation is already done, there is no need for further reasoning traces. Sure they could probably get better alignment with the frontier by distilling fruther, but in any case the damage is already done, at this point hiding reasoning is mostly just hurting users

[epistasis]

It's funny because Anthropic is harming my use of their product, to not even stop the supposed theft of their thought chains. Something that any other supposed upstarts could presumably grab from Moonshot or whatever now.

And my complaint also extends to all their tool use explanations, or rather the lack thereof. I get prompted continually for tool use that I can't examine, that's a poorly formatted 1kb bash script, etc. The PM desire to hide valuable information while requiring extensive interaction has really driven the product into a very unusable place compared to where it was a few months ago. (Or perhaps that's just Claude responding to its memory of my use, and I have somehow driven it to be excessively verbose and difficult to use, which would be unfortunate... Perhaps there's a way to reset the memory.)

[mahkeiro]

The reasoning is key as most of the time the summary provided by fable is not enough to understand the choice and correct the logic. You have to either fully trust it or go to an exhaustive code review. This with the fact that you can only use 4.8 to security review the code produce by fable are the reasons I will not renew my anthropic subscription, the current experience is way to degraded.

[f3408fh]

What will you be replacing it with, if anything?

[gck1]

And recently, since GPT 5.6, OpenAI basically doesn't show anything but a single line, 5 word titles of reasoning traces - titles of summaries of reasoning i presume.

It's effectively just a completely hidden thing now.

[hedora]

Does it have safety guardrails that constantly false positive like Claude does? The only obvious change I’ve seen since opus 4.6 came out is that it constantly flags my requests (no, I’m not doing biology research or security research, yes, it flags for both of those things).

Recently, they backported the blocks to Opus 4.8, so I’m reluctantly stuck on sonnet.

I probably could successfully apply to get special approval to use claude code unencumbered, but I don’t think it is ethical to support tooling that’s built so a central authority gets to decide what intellectual endeavors and knowledge work are permissible, and what are not.

[Deukhoofd]

I feel like the quickstart is missing something. It's referring to its tech blog for actual benchmarks, but K3 isn't mentioned on there, the last thing on that blog was K2.6, 2 releases ago.

[h14h]

> reasoning efficiency matters directly for how expensive a model actually is in real use

I have high hopes on this topic, given token efficiency seemed to be the primary (only?) goal of the K2.7 Code release.

Excited to see the signals that come out of the big eval/benchmark sites.

[cyanydeez]

I eat 1M context in a local model in about 3-4 hours.

It'd need to be exceptionally smart and error free to ever make sense.

[martinald]

Will be interesting to see how it stacks up pricing wise on the various inference providers.

[mmaunder]

Agreed re reasoning. I’ve seen this play out with 5x reasoning negating cost savings.

[schmorptron]

Are thinking models only the reasonable tradeoff vs using much larger non thinking ones because the cost of output tokens is below that of input tokens?

[sroerick]

How do Kimi's subscriptions work? I find their price structure pretty confusing

[fmind-dev]

API prices are amazing, but hosting this on-premise will be real challenge.

[darkbatman]

also its pretty big model inference costs are high even with margins running a 2.8T model costs a lot. if they release oss may be it goes down to $10-12 per million tokens.

[csomar]

It seems the subsidized era is nearing its end and we'll see a convergence on API pricing before a pulling of subscriptions pricing.

[easygenes]

That’s not what this indicates. This is the biggest and most expensive to serve, and most capable open weights model yet. They’re just pricing it in line with capabilities.

Kimi also offers generous subscriptions. Subs aren’t going anywhere. Think of subs like running an insurance business. There might be some users you lose money on (ones who max out their weekly quota without fail), but they’re managed such that the average subscription turns a healthy profit. There’s never been subsidies in model serving, inference is just cheaper in terms of ops TCO than people assume, and API margins are very high.

[csomar]

> They’re just pricing it in line with capabilities.

So... convergence?

> but they’re managed such that the average subscription turns a healthy profit.

It didn't work like that, or at least that's not how it played out. People max-out their subs all the time which is why strict and multiple limits were implemented by all providers. Also, I subscribe to z.ai and recently they dropped the quota significantly that now their sub offers less than Claude and OpenAI. It's still x5-6 what it would cost on API costs though.

> inference is just cheaper in terms of ops TCO than people assume, and API margins are very high.

API margins (at least american ones) are probably healthy. But I don't think that inference is that cheap. It would cost 300-500k to just run GLM 5.2. There are lots of other factors too: reliability (can you keep the GPUs running all time), electricity cost, sys. admin costs, location costs, etc.. I wouldn't be surprised if the API margins are quite close to operational costs.

[ilaksh]

It's as good as gpt 5.6 sol and _half_ the cost..

[nullbio]

Ah, the old "subsidized" meme always rearing its head. Yawn.