Hacker News

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!

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.

The link has 6 well-known benchmarks where this beats Fable (out of 14 I counted). If the numbers hold up scrutiny, this is scary good.

Forget about their pricing but the companies that do have means to host such models fully on-prem are also the same companies that are paying tens of millions of $ in inference cost every month, and are by extension the biggest customers of OAI and Anthropic

I think given how much benchmaxxing we're seeing - the anecdotal evidence of how competent this model is (and efficient) will depend on user's actual real-world use cases.

Given the pricing, it suggests that this model is much more efficient/competent than previous-gen OS/distilled models.

It's like reading Anthropic's obituary.

Crazy how their models always come out after the US labs and just lag the performance of top models. Almost like they are performing distillation attacks... how strange.

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.

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.

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

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.

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.

neither ClosedAI nor Misanthropic will let you use their models without them watching and storing the exchanges indefinitely. no sane company dealing with PII and/or trade secrets allows its employees to use those.

That depends entirely on the hosting situation. If someone can provide a subscription plan at slightly lower rates, it's absolutely compelling.

It’s open weight, so the price will end up being the marginal cost of hosting it.

Personally, I like that there is an option to not send data to companies that have strong financial incentives to steal it.

Also, open weight foundation models can be distilled, so they’re providing a service that the US duopoly is actively blocking. Given that app specific distillation can get > 10x improvements on inference cost (with slight improvement of quality), it’s clear that it’ll win out over time.

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.

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

Charitably, you could read this as "its overall intelligence [is in a class that] ranks second only to [that of]..."

Yeah, bad wording it seems. Though a charitable interpretation is that Fable 5 and GPT 5.6 Sol are joint 1st place in the measurement.

While you are technically correct, in English it’s perfectly fine to say it this way as well.

“Second only” here has meaning “next after”, not “number two”.

Not if the others tie for first place.

Which is still great because it means neither of the two best financed labs in the world manage to produce even two models themselves that would beat Kimi K3.

Sonnet 5 does beat Opus 4.8 on several benchmarks. It just costs more and takes longer.

(On several other benchmarks, it costs more, takes longer, and does worse.)

Possible, but pay-as-you-go Hy3 / DeepSeek v4 Pro / MiMo v2.5 Pro (from respective vendors) are genuinely good enough as daily drivers, given the costs (especially, low prices for input cache, which usually makes up 70%+ of total input for agentic workflows). I put in $10 in DeepSeek & Xiaomi MiMo, and I've barely used $1 each, in a week of coding work.

Coding Plans by MiniMax ($20/mo for 1.7b tokens) and Z.ai (~$30/week use for $17/mo) are also tremendous value for money.

i’ll never really understand this comment. why would labs do this if they know private benchmark evals will come out in the next week?

It was also disruptive because it was open weight, meaning anyone and their dog could theoretically compete with the frontier labs for their inference revenue.

The frontier labs need to recoup a huge amount of cash to cover their model development costs, and justify their valuations. That’s plausible when they’re only ones capable of selling inference on these models, it a lot less plausible when models themselves become cheap commodities, and you’re just competing on your ability to provide compute. Anthropic and OpenAI can’t compete with people like AWS on that front.

cost has nothing to do with why deepseek was disruptive, the fact that it means there is zero moat around anthropic or openai is what's disruptive about it. it means in the mid-term LLMs will be commoditized and customers will flock to the cheapest inference wherever they can find it. there's no reason to stick to the "frontier" labs

DeepSeek didn’t really change any trends though, unless you count the stock market.

It was impressive work, but models were commoditizing and inference costs were dropping rapidly already. They were neither the first nor the last 10x optimization, from what I’ve seen.

It was on the page linked in the top comment, but it's been removed.

I copied that from https://platform.kimi.ai/docs/guide/kimi-k3-quickstart but it seems they updated the page to remove the benchmark score now.

every ai release thread seems to have this same sequence of comments

That's a great question.

I just tried "hi" through the same OpenRouter API and the input token count for that was 86 - and for "hi there" the count was 87.

I think there's an 85 token hidden system prompt of some sort.

I like that it has a snazzy red scarf.

https://en.wikipedia.org/wiki/Mechanical_doping

We don’t know what’s inside these bikes!

The K3 marketing popup when I look at the Kimi Code page says "Kimi K3 Open Frontier Model". So, if it's not going to be open, they haven't told the whole team, yet.

The article says weights will be released in the coming days, and hints it's likely around 50-70B active params.

I've been happilly using kimi models via the $10/month opencode-go[1] subscription for a few months now. I also use pi[2], instead of opencode. Their extensions api is nice, though OpenCode's is similar. My personal preference is more minimalism, add extensions when I want them, instead of the kitchen sink approach.

This is entirely for personal use and small projects. I don't have huge needs. I get access to gpt models via my employer for work things. But I'm also using pi with those models.

[1]: https://opencode.ai/go

[2]: https://pi.dev/

I use everything except for Anthropic's models in opencode.

I don't use Codex CLI myself, but you can configure it to point to OpenRouter instead. OpenRouter has some instructions for Codex CLI and Claude Code here (though they mention Claude Code is not guaranteed to work!):

https://openrouter.ai/docs/cookbook/coding-agents/codex-cli

https://openrouter.ai/docs/cookbook/coding-agents/claude-cod...

Kimi has Kimi Code :)

kimi-code https://www.kimi.com/code/en

I'm so glad to be wrong!

>We are currently working closely with our inference partners and open-source maintainers to align the technical details and ensure the model can be reliably deployed across the ecosystem. The full model weights will be released by July 27, 2026. Further details regarding the architecture, training, and evaluation will be released with the Kimi K3 technical report.

(translated by chrome)

11 days is a long time. It does not take that long to implement inference at providers. In my opinion, seems like they're being pre-emptively cautious about government intervention/review

It came (at least in part) from a document in May where the CCP pretty much said that they will need to review models to make sure they don't threaten national security.

Which basically translates too "Don't give away tools that can be used to undermine your own goals".

It's not even that right now.

I don't understand the question.

The other AIs don't see the question until they are asked to react.

If you read DeepSeek's papers, you'll find a litany of architectural features that allow for a greatly reduced cache hit price by shrinking the size of the KV-cache.

What provider are you using?

> Where did you hear about the deepseek release?

* Tons of gray testing going on for the last 2+ weeks (people at random getting the new v4 model for a while before its removed again).

* It also DeepSeek their 3th birthday this Friday.

* The its been almost 3 months from the v4 DeepSeek release, and the model everybody have been using, was not post-trained. That is what they have been doing during this time.

People trying out the new DSv4 via the web chat with quick game creation tests. People pulling out stuff like Stellaris clones etc.

https://cct124.github.io/HORIZON6_DEMO/

https://www.showyourcode.app/zh/share/pmpwkamrnai2ue

The Battlefront like game is impressive. Sure, the soldiers are backwards and the graphics are still kind of basic. But the entire movement system (run/walk/crouch/jump), gun mechanics, grenades, capture points, AI fighting / capturing back, etc ... Ended up playing it way too darn long lol The text is in mandarin but its not too hard to figure out the menu. Sniper is OP ;)

The Horizon 6 game has everywhere mesh colliders, shows when you off track dirt being kicked up, etc ... In general, both example are very well polished minus the reverse soldiers issue.

And the price is supposed to stay the same (beyond the doubling during Chinese workhours), because everybody got that update.

They emailed current paying users of the api (or at least that’s how I got updated).

This post is at the top when filtered against AI :) Maybe it should use llm based filters to understand if the post is about AI and filter it out?

I'll see your simonw tool and raise you one that actually works: https://hcker.news/?view=frontpage&ai=exclude

I's not just matching against titles. Ironically, I have an agent running daily scans, reading the contents of the top 200 stories of the day. It auto screens high-confidence ones and I make judgement calls on like 10-20 of them per day.

Lol, this post is number one on the leaderboard on the “filtered” list list. Trusting ai slop to filter out ai is as ironic as it gets.

Except it literally shows this post as the first result

How does one get a lobsters invite?

LMArena's "code" leaderboard is really skewed since it's a front-end JS code and design leaderboard. It generates a demo app with two models and then asks "do you prefer A or B". People can look at the code, but most of the time it's just going to be which one looks nicer.

Models that people like the design aesthetic of (Claude, GLM) tend to do better in LMArena than they do on other benchmarks. Design matters, but you look at a model like GPT-5.5 and it's behind Kimi K2.6, Sonnet 4.6, Qwen3.7 Max, and GLM-5.1 on LMArena's code leaderboard. Then you look at benchmarks like DeepSWE and GPT-5.5 blows them out of the water with only Fable and GPT-5.6 beating it.

I'm not saying that the LMArena leaderboard isn't useful, but I'm not sure how much weight I'd give it as a "code" leaderboard. I think often times it's a design comparison of simple front-end React apps rather than a coding comparison. GLM-5.2 is a very good model, but when you look at DeepSWE or Terminal-Bench v2, GPT-5.5 is well ahead.

Odd that open AI models aren't on that graph but are on the rankings! Must be a data lag issue?

TIL, that makes a lot of sense, and thanks for the correction.

Scaling efficiency simply means if you took the first small model and scaled it up to the big model it would take 2.5x the resources to run. Not the that larger model is going to be any cheaper.

Kind of like scaling your personal automobile to the weight of a semi, the semi is still going to be far more efficient in moving cargo, not that the semi will cost the same to operate as the original car.

It's also 2.8x parameter count (1T -> 2.8T), likely higher activation per token (50B?).

The only major player left in this round if I’m not mistaken.

https://www.reuters.com/business/google-gemini-launch-delaye...

most of the gains from the past year and a half have not been from web data, but from synthetic data and agent rollouts with RL.

I dont have estimates on the cost of running models, but I think openai and anthropic are running on subsidized prices. At actual prices it might be worth it in the future.

No, and the reason is simple: Usage is bursty and if you don't maximize usage of the hardware you're going to lose on price.

Ok you can host this model once. What if I want a dozen subagents? Ok you can host it 12 times at once. What if we go a whole week only using max 4 at a time? Etc etc. The limits imposed by self-hosting might be bearable for a variety of reasons, but it's going to be more expensive and less convenient/useful.

There's many obvious excuses ...

Are you claiming a necessity ?

You can limit it a lot to minimize the abuse. In free entrypoint, set token and context limits to be very small. Limit to 2 prompts per IP or something every X hour. That is already a substantial limit where bypassing might not provide much benefits.

Benchmarks look ok, but they don't mention anything about the issue with the model being extremely slow and verbose.

That being said, it's awesome to have such an open-source model, even if now it's unusable mostly locally, with hardware improvements, in a couple of years, the verbosity/speed wouldn't matter as much as the intelligence.

Does that mean this one won't be open source?

USSR is proud to announce that they won 2nd place in an Olympic contest. The filthy USA regime? Next to last!

(There were only two countries competing in said event)

The literal interpretation of that sentence is "when it is second or third, it is only behind Fable 5 or 5.6 Sol". And indeed they give benchmarks where it is ahead of one but not both models.