K2 Horizon: Frontier Performance, Radically Open

38•karimf•1h ago

Comments

piinbinary•16m ago

A bit off topic, but I think I'm starting to get model fatigue. These come out 10x faster than new Javascript frameworks were coming out 10 years ago (at least new models are far easier to adopt).

kelseyfrog•6m ago

Just wait until RSI gains enough traction. We'll be compute-limited rather than labor-limited.

wuhhh•6m ago

At least this one can claim being fully open to differentiate it

kamranjon•16m ago

it's funny that the tagline is Radically Open, but you're immediately hit with http login - maybe this was the wrong link?

gs17•15m ago

https://ifm.ai/k2/ seems to work for me.

esafak•12m ago

But it's missing the all-important charts that the blog had before it started asking for authentication.

wmedrano•11m ago

You can find some of the charts on huggingface

https://huggingface.co/collections/IFM/k2-horizon

sottol•8m ago

Thanks! Qwen-3.8 27B seems to benchmark better but I'd like to try this some time.

sottol•12m ago

It's not the blog post, but there's some info here:

https://ifm.ai/k2/

375 A23B, 36 A4B, 32B, 7B, 3.7B, 0.9B variants.

> 32B: Ranking among the top models in its class, 32B is our most powerful dense model, balancing capability, adaptability, and local deployability.

> 7B: The industry’s best-performing model under 10B combines strong software engineering and expert knowledge in a package small enough to run on a phone.

jjordan•15m ago

Fully open models really need to be a big part of the AI future. That includes all source code, open training data, how it's organized, fed to the model, processed, etc. Until that becomes a thing you're always going to be left wondering what exactly lies underneath the closed model you are using, leaving open the possibility for societal manipulation.

jon9544hn•13m ago

Here’s the link (K2)[https://ifm.ai/k2/] as the originally linked link is a login url.

sottol•11m ago

The press release: https://ifm.ai/k2/press-release/

luckydata•8m ago

both repositories for pre-training and post-training are actually empty... someone might have jumped the gun on the release.

A brief history of learning in imagination

Runtime CSS-in-JS deserves a rehearing

Scientists Find Parasites in Almost Every Fish in a Tuna Haul, 96% Were Infected

Wildwood by Laika Studios

Running modern LLMs in the browser using Three.js and WebGPU

The dark secret behind distressed denim jeans

Yes, we should be worried about the Hugging Face hack

Musk starts spending his midterm election millions, showing key Senate targets

What remained after I tried to automate a real product end-to-end

Show HN: AI Cybersecurity Awareness Game for Cybersecurity Awareness Month

Show HN: Telemetry.dev – OpenTelemetry-native tracing for LLM apps and agents

Show HN: Open-source, Git-inspired versioning for hardware design – Cascadia PLM

MAI-Transcribe-2; fastest, most accurate, and cheapest speech recognition model

Cloudflare is not experiencing a service disruption

The Importance of Personal Archiving

MichiAI 530M parameters full-duplex speech LLM with ~80ms latency

This year I taught myself to code. Today, a tool for everyone, driftcheck

Show HN: TERMy deterministic terminal assistant for Linux

Launch HN: Mireye (YC S26) – Infrastructure for Physical World AI Agents

Steffen's Polyhedron – Greg Egan

Game for a lunch time – ECC Billiards Puzzle

K2 Horizon

Artificial beaver dams saw juvenile coho salmon survival rates go from 8% to 60%

Omacom Foundation hires kernel developer Krzysztof Wilczyński

Trump administration aims to strip schools' tax exemption if they consider race

A Cop's Case for Flock

Show HN: Caption and Cut removes pauses and filler words, burns in captions

Brazil's $97M Discord Lawsuit Could Reshape Big Tech Regulation

IMABO: infinitely many-armed bandit with an oracle for hyperparam optimization

Put-That-There