my ade, plait, is finally in a position to use for my daily work.
github.com/mostlyheath/plait for more
12
LFM3-VL returns its first token on an image question in 200 ms on Qualcomm’s Dragonwing IQ9.
The fastest llama.cpp setup we tested on the same board took 710 ms.
The video shows LFM3-Omni-2B on the same board. How we did it: zetic.ai/blog/lfm3-on-dragon…
We build tailored inference engines for your model on your target hardware: zetic.ai/#contact
Partner program for the ixian app,
Join the waitlist: forms.gle/6KnNYAHvKucBPS1v6
#PhysicalAI #Robotics #AIInference #OnDeviceAI
plait is coming along nicely. it's just for me (for now) but it's open source (MIT) and i'm rolling out updates quickly right now.
2
27
i’ve fallen off on consistency recently, but this is one of the things i’m most proud of. 192 week activity steak on strava.
2
62
heath retweeted
Hiring an inference optimization engineer for your robot?
Meet ixian agent. Give it your model, board, and targets, and it builds the engine.
Our team vs PyTorch eager:
π0.5 on Thor: 9.7× faster
d1-2B on Orin: 4.9× faster
Design partners wanted. DM me your model + board.
5pm? you mean "time for me to walk away from medium screen and stare at small screen while big screen is playing something"?
1
146
it's coming together nicely... lots to add/build. but it's the ADE i've wanted after trying so many of them.
1
44
all this stuff about the magic keyboard being "really good" -- i don't doubt it's good. but it's coming from people who don't use two different macbooks, or N different machines for things.
logitech mx keys mini + mx master is my goat.
1
60
heath retweeted
d1.2-ultra is the #1 model on the Artificial Analysis Decision Index v2.1, with d1.2 placing #6
@liquidai has released d1.2-ultra and d1.2, updating their d1-ultra and d1.1 flagship decision models.
Key results:
➤ d1.2-ultra is the #1 model on AA-Paperclip. With a score of 76, d1.2-ultra is the leading model on AA-Paperclip, our first-party private benchmark that assesses decision-making and extraction capabilities. It leads the next-best model, Drex Max v1.1, by 3 points
➤ d1.2-ultra and d1.2 improve over their predecessors. d1.2-ultra scores 75, a 4-point gain over its predecessor d1-ultra, while d1.2 scores 69, a 3-point improvement over d1.1
➤ d1.2-ultra and d1.2 both sit on the Capability vs. Cost Pareto frontier. d1.2-ultra and d1.2 cost $133 and $63, respectively, per one million Artificial Analysis Decision Index tasks and both models sit on the Capability vs. Cost Pareto frontier
Additional model details:
➤ Context window: 64K tokens
➤ Pricing: $0.042 and $0.02 per 1M input tokens for d1.2-ultra and d1.2, respectively
➤ Multimodality: Text and image input
➤ Providers: Available through Liquid AI’s first-party API
this was SO much fun!!
like i said, we’re putting models on hardware you can’t believe!
Replying to @Qualcomm_Dev
@Qualcomm_Dev 🤝 @liquidai
An evening of LFM3s, powerful models that run locally, and IQ9s, Qualcomm's Dragonwing development boards for running AI right on the device!
🥇 Congrats to team Argus for taking first place and winning IQ9 boards! They built an operating-room assistant prototype that uses vision and audio to track surgical instruments, log spoken updates, and flag missing items. Everything runs locally with LFM3-Omni on an IQ9, keeping patient data on the device.
🥈 Also congrats to team Capy Tower for taking second place and winning Arduino UNO Q boards!
A huge thanks to my cohost, @heath0xFF for putting this together with me! 🦾
Thanks as well to our judges @tugot17, @alex_yuan__, @TianshuYu6, and Charles Coudsi; and to the @zeticai_ team for their Ixian Heagon SDK and support @sj_ondevice and @zetic_ai.
And to everyone who came out to build: thank you for making it such a fun evening 💜
Want to hear about our next events in SF?
Subscribe to our Luma calendar! 👇
luma.com/liquid.ai































