Introducing 𝙻𝙵𝙼𝟹: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence.
> 6 model instances, across text, audio, vision, and omni modalities, generative and decision models.
> 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱: text, 256k context window.
> 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱: vision-language model with video-input support for spatial understanding
> 𝙻𝙵𝙼𝟹-𝙰𝚞𝚍𝚒𝚘-𝟸𝙱: tool calls, voice chat, reasoning in 18 languages.
> 𝙻𝙵𝙼𝟹-𝙾𝚖𝚗𝚒-𝟸𝙱: an omni model combines vision, audio, and language.
> 𝚍𝟷-𝟸𝙱 and 𝚍𝟷-𝙾𝚖𝚗𝚒-𝟸𝙱: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input.
> fine-tune on 1 GPU.
> a new tri-hybrid architecture
> architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs.
> 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters.
> 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱 is #1 on the Pareto frontier of device intelligence 𝚙𝚒𝚙𝚎𝚝𝚝𝚎.𝚕𝚒𝚚𝚞𝚒𝚍.𝚊𝚒
Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device.
1/ 🧵
We just shipped a thumbnail generation feature
And in the next 24 hours we are shipping:
- Niche thumbnail tracker (1d, 3d, 7d, 14d)
- Click test (sharable links)
- Simulator
- 8,000+ manually picked thumbnails for (almost) all niches
- Overpowered MCP
- X-ray analysis
thumbsy.io
One of many features being launched today/tomorrow
phed retweeted
The largest math proof in human history is now computer-verified.
Today, Axiom announces a complete Lean auto-formalization of the Classification of Finite Simple Groups, the "periodic table of symmetry."
About 15,000 pages, built by over 100 mathematicians across generations.
Recommend a feature
And get $100 if we end up adding it
(and 12 months free)
thumbsy.io/suggestions
phed retweeted
I’ve used Frame.io for 5+ years and finally got sick of it.
Introducing post-prod.io ✨
$10 per admin, unlimited free guests, and projects that finish at V2 instead of V8
🖊️Notes on the exact frame
✂️Make or preview changes - Cuts, blur, zooms, subtitles
🏁 Render the final in full quality, no re-export
🔗Link notes to Premiere, upload straight back
Learn more below⤵️
thumbsy.io if you wanna be part of this
At 2,000,000 realtime long form views right now
Long days before the month ends
Next update at 3 mil
Replying to @PhedEU
Only 600k views? Come on Phed, show us your real stats😈 We go back to the Golden years
Running a 3 day sprint for Thumbsy:
- 1 click thumbnail gen
- real time niche thumbnail tracker
- export thumbnail style/brief
- 2,000+ more thumbnails in the pro vault
& much more
the goal was to make the cheapest/useful YouTube tool that you don’t have to think twice about using
thumbsy.io
phed retweeted
Jensen Huang: “… the founders of Higgsfield are visionary in the work that they do … but these are some of the brightest people in the world.”
We're grateful to Jensen and the teams at NVIDIA and TCV for believing in us. Both companies have joined Higgsfield as investors in our Series B
extension.
Since our August Series B:
• We’ve crossed $1 billion in annualized revenue.
• Enterprise contracts are up 71%, growing 100%+ month-over-month
• We’ve grown from 30 million to over 40 million users globally.
TCV has backed Netflix, Facebook, ByteDance and Spotify, companies that changed how people create and consume media. NVIDIA has powered every generation of Higgsfield, from L40 and A40 GPUs through Hopper and Blackwell, with Vera Rubin next to meet growing demand.
Thank you to our creators and customers for building with us every day, and to everyone at Higgsfield whose work brought us here.
The full Bloomberg Announcement is linked below.
Higgsfield, an AI video startup founded by a former Snap executive, said it has attracted investment from Nvidia and TCV to help fund global infrastructure and hiring to capture growing demand. bloomberg.com/news/articles/…
















