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

The Kimi K3 Moment

stephen.bochinski.dev

Post reply on HN
Print view View original

301–310 of 644 posts

Re: The Kimi K3 Moment

#301

Post by tristanj »

3836293648 wrote: post #294
tristanj wrote: post #262

Earlier quoted context omitted.

Partially correct. The court explicitly ruled that training on pirated data, which is what Anthropic was doing, is not considered fair use. Training on legally acquired / licensed data is potentially fair use.

It's not potentially, it's settled. At least for now as neither case wanted to move on to appeals

Not at all. The ruling came from a federal district court, and since it was settled early, it was never reviewed by a higher court. It doesn't set a national precedent across the U.S.

And other district courts don't agree on this. The US district court for Delaware recently rejected a fair use defense for the use of copyrighted works to train AI. https://www.reedsmith.com/articles/court-ai-fair-use-thomson...

There are more cases in the pipeline. The massive NYT vs OpenAI is still ongoing. Nothing will be "settled" until this makes its way to the Supreme Court or Congress steps in.

Re: The Kimi K3 Moment

#302

Post by LUmBULtERA »

credit_guy wrote: post #205

I think it's the opposite. Kimi K3 has 2.8 trillion parameters. We don't know the number of parameters of ChatGPT 5.6 or Opus 4.8, but it's probably in the same region. Fable/Mythos are rumored to be around 10 trillion. So, K3 is directly comparable with ChatGPT 5.6 and Opus 4.8, and the price is not so much lower: K3: $3/$15 per 1 Mtok input/output ChatGPT 5.6 Sol: $5/$30 Opus 4.8: $5/$25 This is not a watershed mom…

Given how OpenAI got rid of their 5-hour limits and reset weekly limits so often, is Kimi really undercutting them on effective price?

Re: The Kimi K3 Moment

#303

Post by fightforcause »

thedreammachine wrote: post #201

It was all distillation up to this point anyway. And I agree with what Suhail said on twitter: "Make the margins next to zero for all these AI models. It was trained on humanity's data, it should be gift to ourselves. Doing so will save us from a few in control of our species."

Normies still thinking this beasts of model coming from China are „dIsTIlLattiOns“ is so funny to me. Many people are not aware of the wave that‘s going to sink US „frontier“ labs that enjoyed and dreamed of stealing tons of data while making people depend on their censored and dumbed down models.

Re: The Kimi K3 Moment

#304

Post by stavarotti »

> I’ve been running Kimi K3 alongside Claude on my normal coding work, and for all practical purposes I can’t tell them apart.

If the author is here, I'm curious what this means. How are they running Kimi K3? Are they using pi, opencode, claude, codex, or kimi-cli? Is speed a concern?

Without knowing how the comparisons are being made, it's hard to agree that one can't notice the difference. I do.

Re: The Kimi K3 Moment

#305

Post by tristanj »

overfeed wrote: post #297
tristanj wrote: post #282

Earlier quoted context omitted.

There's little doubt that Kimi K3 was distilled off Claude. Anthropic stated in February that Moonshot AI (the creator of Kimi) distilled ~3.4 million exchanges from Claude models, as explained in their press release https://www.anthropic.com/news/detecting-and-preventing-dist...

While it sounds like a lot, do you suppose 3.4 million sessions come even close to being sufficient to train a frontier model? Assuming each session was 10,000 words each, that's 34 billion words; lets call it 50 billion tokens (0.05 trillion) unfairly pilfered from Claude. That left Moonshot needing to scrounge for the other 14.950 trillion training tokens required for a baseline frontier model.

3.4 million is the number of sessions Anthropic detected. The actual number of Claude sessions trained on is likely >100 million. There are tens of thousands of accounts funneling Claude sessions into Chinese labs https://www.chinatalk.media/p/how-to-buy-cheap-claude-tokens...

They are used for post-training, i.e. calibrating the model to understand and use tools/command line more effectively.

Re: The Kimi K3 Moment

#306

Post by lebovic »

SwellJoe wrote: post #8

I tried Kimi K3 on a task I've done with every other model I use regularly ( https://swelljoe.com/post/i-let-every-agent-implement-its-ow... ) and found it chewed a lot longer on the problem and ate up almost the entirety of a 5 hour usage limit on their $19 plan. I only have the $20 plan from OpenAI and the same task, with a lot of the same implementation details as Kimi Code, only took a few minutes and consumed al…

Kimi K3 only supports "max" reasoning effort right now, but they plan to enable other levels soon [1].

When I looked at traces from benchmarking, I saw a lot of backtracking and uncertainty while reasoning ("wait, but..."). This also happens with GPT 5.6 and Fable with xhigh/max thinking, albeit to a lesser degree.

I think that explains part of the token inefficiency. Hopefully it will improve with lower reasoning effort settings.

[1]: https://platform.kimi.ai/docs/guide/use-thinking-effort

Re: The Kimi K3 Moment

#307

Post by timmytokyo »

aliasxneo wrote: post #90

Even in this very thread the feedback on Kimi's actual efficacy is debated. I personally feel its worse than both Fable and 5.6 Sol, but I feel like the conversation isn't really about whether its good or not, but a backlash against the U.S governments foray into regulation. So I think people _want_ it to be superior out of anger/frustration with the current situation.

I can't count the number of times I've heard people here say the frontier models are 6 months or more ahead of the open-weights models. That's not true anymore. So the goalposts are shifting.

Re: The Kimi K3 Moment

#308

Post by fibonacci112358 »

ffsm8 wrote: post #160
SwellJoe wrote: post #8

I tried Kimi K3 on a task I've done with every other model I use regularly ( https://swelljoe.com/post/i-let-every-agent-implement-its-ow... ) and found it chewed a lot longer on the problem and ate up almost the entirety of a 5 hour usage limit on their $19 plan. I only have the $20 plan from OpenAI and the same task, with a lot of the same implementation details as Kimi Code, only took a few minutes and consumed al…

Earlier today I made Claude code implement a feature with fable. It worked roughly 60 minutes and used around 30% of my 100€ subs 5h sessions. Then I typed /code-review in a second terminal/clean session after the analysis was done (no code changes) the usage was 99%. I then asked it to write that into a review.md so I could restart from that the next day. Sadly the last % wasn't enough for that. Ymmv, these models b…

/code-review in Claude Code spawns a lot of sub-agents (counted like 8 once), each looking at the code from some certain aspect (like correctness, maintainability, duplication, testing, etc). It eats tokens like crazy doing that, but also covers quite a lot. The default code review in Codex does far less (feels like it's only correctness) and doesn't uses subagents. Actually I made a skill for Codex that does a review closer to what Claude does by default, but using like 4-5 agents and some being cheaper models/less than xhigh reasoning. I'm getting pretty nice reviews with that that cover more than just correctness.

Re: The Kimi K3 Moment

#309

Post by RayVR »

jrflo wrote: post #258
SwellJoe wrote: post #8

I tried Kimi K3 on a task I've done with every other model I use regularly ( https://swelljoe.com/post/i-let-every-agent-implement-its-ow... ) and found it chewed a lot longer on the problem and ate up almost the entirety of a 5 hour usage limit on their $19 plan. I only have the $20 plan from OpenAI and the same task, with a lot of the same implementation details as Kimi Code, only took a few minutes and consumed al…

We really need to stop using $/M tokens as the pricing benchmark. I've found that the number of tokens used tends to be a bigger factor than the listed per token price. The cost per task vs. intelligence curve is really what you care about, and in my estimation Chinese models are just not there. They are focused on benchmaxing and getting the highest raw score they can, rather than efficiency.

Yes, this is already accounted for in many benchmarks, but without deep context of the problem type, the top line pricing is the best starting point.

In my own experience, Fable is more token efficient than opus 4.8 with a higher likelihood of completing tasks correctly or at least with minimal corrective work. Opus regularly struggled to gather the correct context and reason effectively about what it had gathered.

GPT-5.6-sol crushes fable in speed and token efficiency and is clearly superior across many tasks that matter for me.

I also find all models from anthropic after opus 4.6 to suffer from the same ai slop language that long plagued OpenAI and seems to have been reduced drastically in 5.6

Re: The Kimi K3 Moment

#310

Post by timmytokyo »

HarHarVeryFunny wrote: post #269

According to OpenAI's "head of strategic futures": 1) Kimi 3 is a "very good model" 2) It's performance can NOT be explained by distillation 3) The US government should create FUD to stop US corporations from using it (so they use OpenAI instead) https://x.com/deanwball/status/2078133895766114412

Fascinating take from OpenAI. It really gives the lie to the idea that they see AI leading to a better life for all.

"One probable outcome of an open-weight-model-dominant world is full AI communism, which is precisely what China proposes: rather than a market product, AI is a 'public good' which will ultimately be provided by the state as a kind of 'digital public infrastructure.' This future strikes me as a dystopian hellscape, but I've never met an open-weight models advocate who doesn't ultimately concede this is where things end."

He never says why he thinks AI as a "public good" is dystopian, but it's not hard to imagine why. It's because he and his inner circle won't have the power to dictate what we read, see and hear.

Post reply on HN