Cofounder & engineer. Running local AI, building things with it, and sharing what I learn.
Joined April 2022
- Tweets9.5K
- Following771
- Followers8.1K
- Likes16.3K
Kimi K3.1 is out, and it's called Pro 🔥
K3.1 Pro just showed up in the Kimi Code model picker with Low/High/Max thinking, and there's also a K3.1 Pro Highspeed version right under it.
Finally, Kimi K3.1 looks close!
A model called Kibitz showed up on LMSYS Arena today and says it's Kimi. A screenshot from yesterday shows @Kimi_Moonshot Kimi Code tool already naming K3.1 in its model switch warning.
Last time, K3 sat on Arena as Kivine for about 30 hours before it launched in July. Nothing official from Moonshot yet, but if that pattern holds we're talking days.
Glad it's not just me 😂
I've been preferring the voice on my quant over the full model too, and I figured it was just the IKEA effect, liking it more because I made it.
In my thorough testing the new 4B quant by @plotarmordev of Fish Audio's Drama 3 is not only smaller and faster but somehow my custom voices sound better too!
Good work Netrunner!
Let's go 🙏 You've been on the full one, so you're the perfect A/B. Lmk if you hear any difference. Peak should be around 19gb vs ~37 for the full, so you get about 17 GiB back on the Spark
Getting there~ 🫡 you can already run multi models at once with this update. I’ll be trying to trim it more and have a full fledged frontier assistant to fit within 100gb of vram ideally
thanks man 🙏 that's about 17 GiB handed back to the rest of the stack, so the LLM and STT get a lot more breathing room on one Spark
Thanks! I'll be sharing my full setup in depth soon, been polishing it lately.
It started with Hermes on Telegram, where I wanted a voice message with every report. Now I'm trying to move my whole workflow to Mausbot and I still receive a voice recap for every message my assistant sends, so when I'm on the road, busy, or just don't feel like reading a wall of text, I can listen and get the gist.
For prompting I use wisprflow for now, but I'll move to a local model soon enough.
Drama 3 went on a 4bit diet.
My unofficial NVFP4 quant of @fishaudio's Drama 3 drops 28% of its experts and peaks at 19.35 GiB instead of 36.95, with no measurable quality loss in my tests on one DGX Spark.
Capri, my local assistant, explains. Sound on 🔊
Links in the reply:
netrunner retweeted
Issue 3 of AGI Checkpoint is out 🎉
A few highlights for this week 🌟
・@MiniMax_AI put 428B M3.1 Flash weights up at $0.15/$0.60 per million tokens
・@Recursive_SI shipped Flint, a local coding model at
125+ tok/s on a MacBook Pro
・@plotarmordev got TensorFold cloning voices locally
・@ch0mpaa ran TensorFold on a Snapdragon
・@justinn_builds pushed REDSNOW to 275k context on two Sparks
・@Blackwellboy hit 2,008 tok/s prefill with BLACKFOLD
Follow these community builders, they are killing it 🔥
There might be a Kimi K3.1 Pro
A "Kimi K3.1 Pro tech blog" link showed up today in the Research menu on Kimi's site, right above the K3 blog, and then got pulled.
Makes me think Pro is a separate, higher tier above plain K3.1. No date, price or benchmarks yet.