Home Explore Library Help

Search the archive⌘K

A living room for X videos. Grab, watch, keep.

PRODUCT

Explore
Library
Settings

LEARN

Field notes
Help & FAQ

LEGAL

Terms of Use
Privacy Policy
Report content

© 2026 XDL · Built for the den

Kimi-K3 just got 62% compressed to run locally with 79% accuracy intact. · XDL

Home
Explore
Kimi-K3 just got 62% compressed to run locally w…

Kimi-K3 just got 62% compressed to run locally with 79% accuracy intact.

@Hugging Models

2 grabs

Up next

TUNED TO THIS CLIP

Meet Mistral Large 4, aka Le Chonk. • 1T parameters, natively multimodal. 49B active. It is the best open weights model from US or Europe on aggregated benchmarks. • State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses

Minimax H3 running locally. ~65GB of models. 12GB VRAM. RTX 4070 Ti. 100% Text-to-Video. 100% ComfyUI. Pass 1 → 960×544 Pass 2 → 1920×1088 One continuous loop built from multiple context-connected generations. No cloud. No API. CONTROL IS THE KEY. LOCAL IS THE WAY.

@Maurice Bourdon

Kimi K3 vs Fable 5. Same prompt and reference. 3D globe dashboard. Kimi left, Fable right. To me, Kimi's UI fidelity is a touch better. The globe looks better too, with richer textures and smoother motion. This task: ~$3 with Kimi, ~$15 with Fable. I'd pick Kimi k3 for

Here is a similar one I made with Kimi K3

Kimi K3 just modeled a $31 million hotel, all the piping, wiring, ductwork and fire protection systems for 190 rooms, fully coordinated in advance. Before: 3 engineers, 6 weeks, $47,000. Now: 1 engineer reviewing AI output, 9 days, $10,500. Kimi connected directly to the

I've tried out Kling 4.0 flash and it didn't disappoint! Fast movement has seen a huge improvement compared to Kling 3.0! Great work @Kling_ai

GPT-6 Astra scored 95% on a robot control task, up from Fable 5.1's 40%, with 6.2x fewer output tokens at 2.3x lower cost. 🧵

No comments yet

Write a comment

Sign in with X to post

Say something. You’re first.