Live data from Hacker News

Kimi K3-256k

kimi.com

71–80 of 172 posts

Re: Kimi K3-256k

#71

Earlier quoted context omitted.

I'm always curious what you guys are working on; every git repo I've run a local model on and stick below <100k to increase speed seems effective enough to scope patches and changes.

My current Claude Code session has been going on for like 35 hours and has used up around 400 million tokens, thankfully almost all of those being cached (95-98%) - pretty typical for long form agentic work. First you spend like 2-3 hours working on a plan, once you have that you just tell the model to go and implement it, do adversarial sub-agent review loops before each commit and also make sure that all tooling an…

Kimi CLI has a mechanism with checkpoints and the ability for the agent to revert to a checkpoint + a message of how to continue based on what went right/wrong. I don't know if that's the cause of what you've seen, but it's plausible.

Re: Kimi K3-256k

#72

Wow. So kimi is suddenly half the price for all users until they hit 256k of context? Thats massive.

I don't think so. This is a separate model, so I assume that if you just use this and switch to the 1 million context model when you reach 256k, your cache will be invalidated, so you'll re-pay the 256k tokens on the 1 million context model pricing. Edit: I was wrong, thanks to longwave for pointing this out. It's absolutely possible to start out on the 256k model and then switch to the 1 million model when you get c…

The article explicitly says "The current version switching from 256k to 1M does not affect the cache."

Re: Kimi K3-256k

#73
post #61

LLMs is quickly became commodities. US AI labs like OpenAI is losing their moat. Hyperscalers and data center owners who can sell cheap token will win

I believe Codex harness is pretty sticky though I haven’t tried many others. Does anyone else provide a harness of that quality?

Pi harness with subagents is pretty great. I’m using it for everything now.

Re: Kimi K3-256k

#74
post #2

[flagged]

This exact same comment is the top comment on the reddit thread for this news https://www.reddit.com/r/kimi/s/BFa1TR9vNg

Does that URL link back to you somehow?

Full: https://reddit.com/r/kimi/comments/1v9aqsi/comment/p0cbe0n/

Mirror: https://redlib.us.catsarch.com/r/kimi/comments/1v9aqsi/k3256...

Re: Kimi K3-256k

#75
post #29

Earlier quoted context omitted.

Anthropic services have outages on all days ending in y.

So users in Germany are not affected?

Anthropic has better SLA in Germany. I’ve heard uptime there can get up to nein nines.

Re: Kimi K3-256k

#76

Earlier quoted context omitted.

It isn't legally possible for them to do this at the EU level. The EU parliament would never vote for it. For pressure at the country level leading to this kind of thing I think it's very unlikely. Here in Sweden it wouldn't just require a vote in the Swedish parliament and before this there'd have to be förarbeten and you can't just brazenly push things through with insane arguments, Swedish social convention goes a…

It happened with ASML. What's your thought on it? The US forbid dutch company to export.

Well, that's the deal, I assume-- that they weren't allowed to buy the Japanese light source outright, so they bought the American one, even though it required giving the Americans some sort of veto or control.

I guess it sucks if one wants to expert broadly, but if you're big on vertical integration and the Japanese won't sell I guess you take what you can get.

Re: Kimi K3-256k

#77

Earlier quoted context omitted.

Once you have the plan you don't need to keep the 2hrs of research in the context (which is most of it) you can drop that plan into a file and start fresh for implementation.

That is true, but I feel like sometimes if the conversation contains useful rational it can help to keep it. I think sometimes it is a judgement call, I will sometimes compress the context first. If I feel like the model and I explored a lot of options I won't want to keep context as it might be confusing. I think the more you use it the better judge you are of whether you should purge, compress, or just keep the con…

Models work best when they have short instructions and no noise. "Conversation [may] contain" also means "conversation has a lot of noise". It degrades performance and increases cost.

I use handoff skill to ask model to write a prompt for itself.

Re: Kimi K3-256k

#78

What is the purpose of this? Just a hard cutoff below the actual context window? You could set that in your harness anyway.

At least in the self-hosted LLM inference engines, you have to pre-allocate space for the maximum amount of context you want to allow for each parallel session. By using a lower maximum, you don't have to allocate as much VRAM for each session, allowing more usage for the same amount of hardware. Thus, cheaper.

Re: Kimi K3-256k

#79
post #15

Earlier quoted context omitted.

Since the model is open-weights, you can get from other providers, for example see https://openrouter.ai/moonshotai/kimi-k3#providers

The only downside with third party providers is that you have to trust that the provider have setup and configured it correctly, and is not secretly quantizing it. See Kimi Vendor Verifier.

Plenty of providers on that list aren't fly-by-nights and have easily proven themselves in past models. Jeremy Howard does demos on Fireworks.

A nice thing about open router is you can specify filters, like "US hosting without data retention"

Re: Kimi K3-256k

#80
post #71

Earlier quoted context omitted.

My current Claude Code session has been going on for like 35 hours and has used up around 400 million tokens, thankfully almost all of those being cached (95-98%) - pretty typical for long form agentic work. First you spend like 2-3 hours working on a plan, once you have that you just tell the model to go and implement it, do adversarial sub-agent review loops before each commit and also make sure that all tooling an…

Kimi CLI has a mechanism with checkpoints and the ability for the agent to revert to a checkpoint + a message of how to continue based on what went right/wrong. I don't know if that's the cause of what you've seen, but it's plausible.

With prefix caching, you get checkpoints for free. Do you mean that?
Post reply on HN