The launch
Liquid AI posted at 17:07 UTC on 27 October 2026. One post, six models and a 2:17 film.
LFM3 is the new generation of Liquid Foundation Models. It runs on a new tri-hybrid architecture with new multimodal encoders, across text, audio, vision and omni.
Liquid AI spun out of MIT CSAIL. Its founders are Ramin Hasani, Mathias Lechner, Alexander Amini and Daniela Rus.
Introducing LFM3: our new generation of Liquid Foundation Models.
The post leads with the name and the generation, then lists the six models. A builder can tell in the first seconds whether it is for them.
Past NVIDIA and Mistral
Both launched the week before, from larger accounts. NVIDIA posted Nemotron 3.5 Bolt to 356K followers. Mistral posted Voxtral 3 to 226K.
LFM3 was within a few thousand views of both at 95 minutes. By hour five it had more than twice the views of either.
Launches from the large labs arrive with an audience already waiting. LFM3 had to find its own, and found more of one in an afternoon.
The first day, reading by reading
The directory read the post through its first day. These are the readings, with the time since it went out.
| After | Views | Bookmarks | Quotes |
|---|---|---|---|
| 52 min | 121K | 518 | 110 |
| 1 h 35 | 268K | 900 | 167 |
| 4 h 53 | 637K | 2,386 | 299 |
| 5 h 51 | 742K | 2,929 | 330 |
| 11 h 13 | 1.09M | 5,358 | 496 |
Builders saved it. Bookmarks nearly doubled between hours six and eleven, from 2,929 to 5,358, while views grew by about half.
The post stands at 48.9 saves per 10,000 views. The directory median is under 12.
The quotes came early. At 52 minutes there were 110 quote posts against 995 likes, one quote for every nine likes. The directory median is one for every twenty.
Against its own last launch
The fairest measure of a launch is the company’s own last one. On 7 October Liquid AI released Open d1, two open-weight multimodal models, from the same account.
| Launch | Views | Bookmarks |
|---|---|---|
| LFM3, 27 Oct | 1.1M | 5,358 |
| Open d1, 7 Oct | 433K | 1,568 |
LFM3 passed Open d1’s lifetime total inside five hours. Same account, same audience, three weeks apart. Doomers ran the second one.
What Doomers ran
Doomers ran the launch window. Researchers, engineers, founders and creators from the network, with audiences that build with models, quoted the post in the first hour with their own take on the architecture.
That first wave took a company account of 40,000 into the feeds of people who build on open and on-device models.
The account gained about 3,000 followers in eleven hours.
LFM3 by the numbers
| Measure | Value |
|---|---|
| Views | 1,094,756 |
| Bookmarks | 5,358 |
| Saves per 10,000 views | 48.9, against a directory median under 12 |
| Quote posts | 496 |
| Followers | about 40,000 at launch, 43,020 now |
The post
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
See the live numbers on the launch page →
Questions we get asked
What is LFM3?
LFM3 is Liquid AI’s new generation of Liquid Foundation Models, introduced on 27 October 2026: six models across text, audio, vision and omni, on a new tri-hybrid architecture with new multimodal encoders.
Who is behind Liquid AI?
Liquid AI is a spin-out of MIT CSAIL, founded by Ramin Hasani, Mathias Lechner, Alexander Amini and Daniela Rus.
Who ran the LFM3 launch?
Doomers. Researchers, engineers, founders and creators from the network, with audiences that build with models, quoted the post in the first hour with their own take.
How many views did the LFM3 launch get?
The launch post stands at 1.1M views and 5,358 bookmarks, from a company account of about 40,000 followers. Inside five hours it passed NVIDIA’s Nemotron 3.5 Bolt and Mistral’s Voxtral 3 launches, and Liquid AI’s own Open d1 release from three weeks earlier.