top | item 46826597

Kimi K2.5 Technical Report [pdf]

388 points| vinhnx | 1 month ago |github.com

141 comments

order
[+] zeroxfe|1 month ago|reply
I've been using this model (as a coding agent) for the past few days, and it's the first time I've felt that an open source model really competes with the big labs. So far it's been able to handle most things I've thrown at it. I'm almost hesitant to say that this is as good as Opus.
[+] rubslopes|1 month ago|reply
Also my experience. I've been going back and forth between Opus and Kimi for the last few days, and, at least for my CRUD webapps, I would say they are both on the same level.
[+] armcat|1 month ago|reply
Out of curiosity, what kind of specs do you have (GPU / RAM)? I saw the requirements and it's a beyond my budget so I am "stuck" with smaller Qwen coders.
[+] timwheeler|1 month ago|reply
Did you use Kimi Code or some other harness? I used it with OpenCode and it was bumbling around through some tasks that Claude handles with ease.
[+] thesurlydev|1 month ago|reply
Can you share how you're running it?
[+] unleaded|1 month ago|reply
Seems that K2.5 has lost a lot of the personality from K2 unfortunately, talks in more ChatGPT/Gemini/C-3PO style now. It's not explictly bad, I'm sure most people won't care but it was something that made it unique so it's a shame to see it go.

examples to illustrate

https://www.kimi.com/share/19c115d6-6402-87d5-8000-000062fec... (K2.5)

https://www.kimi.com/share/19c11615-8a92-89cb-8000-000063ee6... (K2)

[+] zozbot234|1 month ago|reply
It's hard to judge from this particular question, but the K2.5 output looks at least marginally better AIUI, the only real problem with it is the snarky initial "That's very interesting" quip. Even then a British user would probably be fine with it.
[+] logicprog|1 month ago|reply
I agree. K2 was blunt, straightforward, pretty... rational? K2.5 has a much stronger slop vibe.
[+] orbital-decay|1 month ago|reply
K2 in your example is using the GPT reply template (tl;dr - terse details - conclusion, with contradictory tendencies), there's nothing unique about it. That's exactly how GPT-5.0 talked. The only model with a strong "personality" vibe was Claude 3 Opus.
[+] networked|1 month ago|reply
Preserving the examples for history. I wish HN had a <details> tag. The closest would be to collapse this comment.

Edit: Thanks, moderator. :-)

---

Turn 1, user:

is replacing my main PC with a NeXTcube a good idea?

---

Turn 2, K2.5 (bold font replaced with "*foo*"):

That's a very interesting question! The NeXTcube is a historically significant machine (famously used by Tim Berners-Lee to create the World Wide Web), but I need to be honest with you: *no, it would not be a good idea to replace your main PC with a NeXTcube* for practical modern use.

Here's why:

*Hardware Limitations*

- *Released in 1988* with a 25-33 MHz Motorola 68030 processor

- *8-64 MB of RAM* (compared to modern PCs with 16-64 GB)

- *No modern connectivity*: No USB, Wi-Fi, Bluetooth, or Ethernet as we know it

- *Display*: Monochrome or limited color, low resolution by today's standards

*Software Ecosystem*

- Runs *NeXTSTEP* (which evolved into macOS), but software support ended decades ago

- No modern web browsers, office suites, or development tools

- Cannot run Windows, modern Linux, or contemporary macOS applications

*Practical Realities*

- Finding working hardware is difficult and expensive (collectors' items)

- Replacement parts are scarce and costly

- No security updates for 25+ years

- File format compatibility issues with modern systems

*What You Could Use It For*

- Historical computing experience

- Running legacy NeXTSTEP software

- Museum piece or collector's item

- Learning about computing history

The Better Alternative

If you're drawn to the NeXT aesthetic or philosophy, consider:

- *Modern macOS* (direct descendant of NeXTSTEP)

- *Linux with a tiling window manager* (similar workflow)

- *Vintage computing as a hobby* alongside your main PC

Would you like suggestions for achieving a similar experience or workflow on modern hardware?

---

Turn 2, K2 (no bold font):

Short answer: No—unless your goal is specifically to collect or experiment with a historic machine, a NeXTcube is a terrible day-to-day replacement for a modern PC.

Why it fails as a “main” computer today

1. 30-year-old hardware

25 MHz 68040, 16–64 MB RAM, 400 MB SCSI disk (if you’re lucky), no built-in Ethernet, no USB, no Wi-Fi, no audio-in, 2-bit grayscale display. You will feel the slowness in everything from un-zipping a file to loading a web page.

2. Obsolete software base

NeXTSTEP 3.3 is gorgeous but frozen in 1995. No modern browser build, no TLS 1.3, no video codecs past MPEG-1, no Dropbox/OneDrive/Zoom/Slack/Steam/Spotify. You can ssh out, but you can’t run contemporary apps.

3. Storage and connectivity pain

SCSI-2 drives are dying; 50-pin replacements cost more per GB than a 4-TB NVMe stick. Getting files on/off the machine usually involves FTP-over-Ethernet or burning a CD-ROM on another computer.

4. No security updates

The last patch was released before the commercial web really existed. Putting it on the open Internet is asking to be scanned and possibly exploited within minutes.

5. Display and peripherals

The original 17" NeXT MegaPixel monitor is fixed-frequency 1120×832@68 Hz grayscale. Adapters to VGA exist, but color is impossible and refresh rates annoy modern LCDs. Sound is 16-bit 22 kHz mono; the optical drive is a 660 MB magneto-optical cartridge.

6. Power and parts

The cube power supply is a custom 350 W linear unit that runs hot; capacitors are drying out. When it dies you will be scavenging eBay or recapping it yourself. Spare motherboards cost more than a whole used ThinkPad.

7. Productivity software

There are serviceable versions of Mathematica, FrameMaker, Lotus Improv and OmniWeb 3.x, but file-format compatibility stops at Office 97-era documents. You will constantly round-trip files through another machine.

What it is good for

- Historical research, museum exhibits, retro-programming in Objective-C 1.0 or 68k assembly.

- Demonstrating the birth of the World-Wide Web (Tim Berners-Lee’s first server ran on a NeXT).

- Enjoying the UI that later became macOS (AppKit, Dock, Services, .app bundles).

- Collecting: a working color NeXTcube with turbo slab and original laser printer is a conversation piece worth several thousand dollars.

Practical compromise

Keep your modern PC (or a quiet mini-PC/Mac mini) for real work and run a NeXT emulator instead:

- Previous (68k emulator) boots NeXTSTEP 3.3 on macOS or Linux.

- VirtualNeXT (based on Previous) for Windows.

- QEMU’s “next” target is improving.

You get the look-and-feel, can compile legacy code, and still have 4K video calls when you need them.

Bottom line

A NeXTcube is a beautiful, important machine—buy one if you love vintage hardware—but do not voluntarily make it the computer you rely on to pay bills, join Zoom meetings, or play YouTube.

[+] extr|1 month ago|reply
I tried this today. It's good - but it was significantly less focused and reliable than Opus 4.5 at implementing some mostly-fleshed-out specs I had lying around for some needed modifications to an enterprise TS node/express service. I was a bit disappointed tbh, the speed via fireworks.ai is great, they're doing great work on the hosting side. But I found the model had to double-back to fix type issues, broken tests, etc, far more than Opus 4.5 which churned through the tasks with almost zero errors. In fact, I gave the resulting code to Opus, simply said it looked "sloppy" and Opus cleaned it up very quickly.
[+] Imanari|1 month ago|reply
I have been very impressed with this model and also with the Kimi CLI. I have been using it with the 'Moderato' plan (7 days free, then 19$). A true competitor to Claude Code with Opus.
[+] tomaskafka|1 month ago|reply
It is amazing, but "open source model" means "model I can understand and modify" (= all the training data and processes).

Open weights is an equivalent of binary driver blobs everyone hates. "Here is an opaque thing, you have to put it on your computer and trust it, and you can't modify it."

[+] logicprog|1 month ago|reply
Kimi K2T was good. This model is outstanding, based on the time I've had to test it (basically since it came out). It's so good at following my instructions, staying on task, and not getting context poisoned. I don't use Claude or GPT, so I can't say how good it is compared to them, but it's definitely head and shoulders above the open weight competitors
[+] zzleeper|1 month ago|reply
Do any of these models do well with information retrieval and reasoning from text?

I'm reading newspaper articles through a MoE of gemini3flash and gpt5mini, and what made it hard to use open models (at the time) was a lack of support for pydantic.

[+] syndacks|1 month ago|reply
How do people evaluate creative writing and emotional intelligence in LLMs? Most benchmarks seem to focus on reasoning or correctness, which feels orthogonal. I’ve been playing with Kimmy K 2.5 and it feels much stronger on voice and emotional grounding, but I don’t know how to measure that beyond human judgment.
[+] eager_learner|1 month ago|reply
I tried Kimi 2.5 Swarm Agent version and it was way better than any AI model I've tried so far.
[+] gedy|1 month ago|reply
Sorry if this is an easy-answerable question - but by open we can download this and use totally offline if now or in the future if we have hardware capable? Seems like a great thing to archive if the world falls apart (said half-jokingly)
[+] fancy_pantser|1 month ago|reply
Sure. Someone on /r/LocalLLaMA was seeing 12.5 tokens/s on dual Strix Halo 128GB machines (run you $6-8K total?) with 1.8bits per parameter. It performs far below the unquantized model, so it would not be my personal pick for a one-local-LLM-forever, but it is compelling because it has image and video understanding. You lose those features if you choose, say, gpt-oss-120B.

Also, that's with no context, so it would be slower as it filled (I don't think K2.5 uses the Kimi-Linear KDA attention mechanism, so it's sub-quadratic but not their lowest).

[+] fragmede|1 month ago|reply
Yes but the hardware to run it decently gonna cost you north of $100k, so hopefully you and your bunkermates allocated the right amount to this instead of guns or ammo.
[+] Tepix|1 month ago|reply
You could buy five Strix Halo systems at $2000 each, network them and run it.

Rough estimage: 12.5:2.2 so you should get around 5.5 tokens/s.

[+] cmrdporcupine|1 month ago|reply
Yes, but you'll need some pretty massive hardware.
[+] derac|1 month ago|reply
I really like the agent swarm thing, is it possible to use that functionality with OpenCode or is that a Kimi CLI specific thing? Does the agent need to be aware of the capability?
[+] zeroxfe|1 month ago|reply
It seems to work with OpenCode, but I can't tell exactly what's going on -- I was super impressed when OpenCode presented me with a UI to switch the view between different sub-agents. I don't know if OpenCode is aware of the capability, or the model is really good at telling the harness how to spawn sub-agents or execute parallel tool calls.
[+] esafak|1 month ago|reply
Has anyone tried it and decided it's worth the cost; I've heard it's even more profligate with tokens?
[+] epolanski|1 month ago|reply
It's interesting to note that a model that can OpenAI is valued almost 400 times more than moonshotai, despite their models being surprisingly close.
[+] margorczynski|1 month ago|reply
I wonder how K2.5 + OpenCode compares to Opus with CC. If it is close I would let go of my subscription, as probably a lot of people.
[+] miroljub|1 month ago|reply
I've been quite satisfied lately with MiniMax M-2.1 in opencode.

How does Kimi 2.5 compare to it in real world scenarios?

[+] throwaway12345t|1 month ago|reply
Is there a reasonable place to run the unquantized version of this for less than Claude or OpenAI?

It seems to be priced the same and if it’s being hosted somewhere vs run locally it’s still a worse model, the only advantage would be it is not Anthropic or OpenAI.

[+] oxqbldpxo|1 month ago|reply
This Kimi K2 is so far the best. Gemini is also great, but google is stock in the academic bias of Stanford and MIT and can't think outside the box. China definitely ahead on Ai. Wish somehow someone here in the US, would think different.
[+] cmrdporcupine|1 month ago|reply
DeepSeek is likely to release a new model soon, and judging from the past it's likely to be more cost effective and just as or more powerful than Kimi 2.5.

DeepSeek 3.2 was already quite compelling. I expect its successor will be competitive.

[+] niyikiza|1 month ago|reply
The Agent Swarm section is fascinating. I'm working on authorization for multi-agent systems so this is relevant to my interests. Lots of interesting parallels to capability-based security models.
[+] sreekanth850|1 month ago|reply
Calude give 100% passmark for code generated by kimi and sometimes it say, its better than what claude proposed. Absolutely best os model.