Joined November 2020
- Tweets56
- Following223
- Followers275
- Likes333
Today we release 𝚍𝟷-𝚞𝚕𝚝𝚛𝚊, our most powerful decision model.
> built for hard decision problems.
> d1-ultra is multimodal
> supports 44 languages
> #1 on the @ArtificialAnlys Decision Index (65)
> leads all eight categories of our 112-evaluation suite
1/
Tim Seyde retweeted
Today we release 𝚍𝟷-𝚞𝚕𝚝𝚛𝚊, our most powerful decision model.
> built for hard decision problems.
> d1-ultra is multimodal
> supports 44 languages
> #1 on the @ArtificialAnlys Decision Index (65)
> leads all eight categories of our 112-evaluation suite
1/
Our latest 24B MoE is here! Try it for yourself:
huggingface.co/LiquidAI/LFM2…
Today we release 𝙻𝙵𝙼𝟸.𝟻-𝟸𝟺𝙱-𝙰𝟸𝙱, the largest 𝙻𝙵𝙼𝟸.𝟻 family member. It is an efficient local model for laptops and PCs to work through multi-step tasks inside popular agent harnesses.
> 24B total parameters, 2B active per token
> Context expanded to 131K tokens
> A 128K-token vocabulary
> 33T pretraining tokens
> It is a reasoning model
> Very easy to fine-tune and customize
> Performs competitively with Qwen3.6-35B-A3B, a model with ~50% more total and active parameters, on agentic and tool-use tasks.
> Open base and post-trained weights on @huggingface 🤗
1/
Tim Seyde retweeted
We're partnering with @Qualcomm to bring personal AI context to devices powered by Snapdragon. Liquid Context, our on-device context layer, is now optimized for Snapdragon processors and runs on the Qualcomm Hexagon NPU.
Our goal: Give the agents people choose an understanding of what matters to them and when they need help. With the user's permission, Liquid Context learns from device signals and builds an understanding of their routines, preferences, and needs. That understanding is built and maintained locally, and Liquid Context shares relevant context with the user's chosen agents, whether they run on the device, in the cloud, or across both. That includes third-party agents and Liquid Agent, our efficient embedded agent powered by LFM2.5-2.6B.
Running on the Hexagon NPU, Liquid Context works in the background and keeps that understanding current without requiring a cloud model to process every update.
For device manufacturers, this is a path to add personal context to their devices while supporting their own choice of agents and services. OEMs building embedded or hybrid agents can also work with us to evaluate Liquid Agent.
As our CEO @ramin_m_h said: "Personal AI starts with understanding how you live and what you need, when you need it. Liquid Context builds that understanding on your device so the agents you choose can offer more relevant help and anticipate your needs."
> Read more about our partnership: liquid.ai/blog/liquid-contex…
> Check out the livestream of @cristianoamon and Ramin's keynote here: youtube.com/watch?v=xKCto1Yf…
Tim Seyde retweeted
Today we release Pipette, a model evaluation suite for on-device intelligence, in partnership with @ArtificialAnlys
Most benchmarking platforms are optimized to measure core capabilities and speed profile of foundation models served in the cloud. Pipette gives the field a common, reproducible way to measure the quality, speed, latency, and memory use of AI models on devices such as phones, laptops, PCs, AI boxes, and embedded hardware.
> Pipette is open source
> In Pipette, models get compated as model + quantization + runtime + device from one interface.
> It comes with a warehouse of verified benchmark results, currently with 10k+ results across 35 model classes, 7 quants, llama.cpp runtimes, and 4 devices.
> Pipette is a dynamic platform, allowing new contributions from day one, adding new devices, runtimes, model families, and quantization levels.
🧵 – at San Francisco, CA
Tim Seyde retweeted
Today, we release LFM2.5-VL-3B, a lightweight vision-language model that reads screens, documents, and the physical world. It handles digital screens across mobile, web, and desktop, grounds objects to coordinates, reads text and charts, and calls tools from either text or image input.
Built on LFM2.5-2.6B base, with a SigLIP2 400M NaFlex vision encoder
> Pre-trained on ~34T tokens
> Vocab size: 128K
Comparable or better scores compared to models up to 2.6x its size:
> ScreenSpot-v2 80.7, ahead of Gemma-4-E4B at 51.2
> RealWorldQA 73.1, ahead of InternVL-3.5-4B at 67.7
> TextVQA 84.3, ahead of Qwen3.5-4B at 81.2
> RefCOCO-avg 87.9, up from 57.1 on LFM2-VL-3B
> ToolSandbox 59.5, up from 26.4 on LFM2-VL-3B
🧵
Excited to share LFM2.5-2.6B: small, fast, and built for local agentic workflows. Try it out!
Today we release LFM2.5-2.6B, an agentic model that runs entirely on-device. It plans, calls tools, and works through multi-step tasks on phones, laptops, PCs, and robots. Data never leaves the device, and the marginal cost of each run is essentially zero.
> Pre-trained on ~34T tokens
> LFM2.5 flagship hybrid architecture
> Context length: 128K
> Vocab size: 128K
> balanced intelligence per watt
> customizable on a single GPU for any specialized task
> LFM2 open-weight license
Comparable or better scores compared to models up to nearly 4x its size:
> ToolSandbox 77.83, ahead of Qwen3.5-9B at 76.44
> Multi-IF 80.07, ahead of Gemma-4-E4B-it at 77.35
> IFStruct 85.49, ahead of Qwen3.5-9B at 78.50
🧵
Dumbo's first steps — LFM2.5-230M doing multi-step tool-calling over pre-trained skills provided by @nvidia SONIC. Same small model, many different use cases.
As an early look at ongoing work, we deployed LFM2.5-230M on a Unitree G1, running entirely on-device on its onboard @nvidia Jetson Orin.
The model acts as a skill-selection layer, taking in natural-language instructions and decomposing them into sequences of tool calls.
After a quick fine-tune, "Hold still for 2s, walk forward at 1 m/s for 3 m, hold a one-leg kneel for 5s, walk back at 0.5 m/s for 3 m" becomes a structured multi-step plan automatically.
(3/n)
Tim Seyde retweeted
Introducing LFM2.5-230M: our smallest model yet, built to run fast anywhere (CPUs, NPUs, and GPUs) to enable agentic tasks on phones, robots, home and network automation devices.
> 230M parameters, built on the LFM2 architecture
> Pre-trained on 19T tokens, with a 32K context extension
> Post-trained with distillation from LFM2.5-350M
> 213 tok/s decode speed on Galaxy S25 Ultra (CPU)
> 42 tok/s on a Raspberry Pi 5 (CPU)
> Competes with and often beats models more than twice its size on instruction following, data extraction, and tool use.
> use it for large-scale data extraction pipelines or lightweight on-device agentic workloads.
🧵
Tim Seyde retweeted
Today, we're releasing LFM2.5-8B-A1B, a device-optimized model designed to power real-life applications on phones, laptops, PCs, robots, and fast & lightweight server-side use-cases.
> 8B MoE, 1.5B active
> Expanded 128K context
> LFM2.5 flagship hybrid MoE architecture
> Trained on 38T tokens + large-scale RL
> fast, reliable tool calling, punching above its weight, comparable to models with up to 4x its size
> customizable on a single GPU for any specialized task
> LFM2 open-weight license
🧵
Honored to have curated the data for this
Three years ago we started working on a stealth project that we weren’t sure we’d ever talk about publicly... until today.
Breakthrough: Introducing LFM-Zero: the first foundation model trained on 0 tokens.
No pretraining. No finetuning. No data. Instead, we initialize from an implicit probabilistic prior over the underlying data-generating process, allowing the model to converge without ever observing data.
LFM-Zero matches or surpasses models trained on 10T+ tokens across reasoning, coding, and multimodal tasks. Turns out that pretraining was just regularization that was holding us back.
> Read our Tech Report here: tinyurl.com/lfm-zero
Our smallest model just got better — try it out!
Tim Seyde retweeted
Today, we release our largest LFM2 model: LFM2-24B-A2B 🐘
> 24B total parameters
> 2.3B active per token
> Built on our hybrid, hardware-aware LFM2 architecture
It combines LFM2’s fast, memory-efficient design with a Mixture of Experts setup, so only 2.3B parameters activate each run.
The result: best-in-class efficiency, fast edge inference, and predictable log-linear scaling all in a 32GB, 2B-active MoE footprint.
🧵
We’ve released LFM2.5-1.2B-Thinking on @huggingface.
It’s a fast reasoning model that runs on-device. Enjoy!
huggingface.co/LiquidAI/LFM2…
Today we release LFM2.5-1.2B-Thinking, a reasoning model that runs entirely on-device. What needed a data center two years ago now runs on any phone with 900 MB of memory.
> Trained specifically for concise reasoning
> Generates internal thinking traces before producing answers
> Enables systematic problem-solving at edge-scale latency
> Shines on tool use, math, and instruction following
Tim Seyde retweeted
Together with @maximelabonne , Alfred Eng, and @timseyde, I trained LFM2-2.6B-Transcript, a private, on-device meeting summarization model released with AMD.
Cloud-quality summaries in seconds, <3GB RAM, fully local on Ryzen™ AI PCs.
huggingface.co/LiquidAI/LFM2…
We are releasing our new LFM2.5-1.2B family of models.
Try them here:
Text: huggingface.co/collections/L…
Vision: huggingface.co/collections/L…
Audio: huggingface.co/collections/L…
Today, we release LFM2.5, our most capable family of tiny on-device foundation models.
It’s built to power reliable on-device agentic applications: higher quality, lower latency, and broader modality support in the ~1B parameter class.
> LFM2.5 builds on our LFM2 device-optimized hybrid architecture
> Pretraining scaled from 10T → 28T tokens
> Expanded reinforcement learning post-training
> Higher ceilings for instruction following
🧵
LLMs that predict their own response quality can allocate test-time compute more efficiently. ZIP-RC repurposes unused logits to model joint distributions over reward and generation length. At every token, no architecture changes, no extra forward passes - great for adaptive test-time sampling.
Exciting work led by @rohin_manvi & amazing team across @liquidai, @berkeley_ai, @MIT_CSAIL.
🚨Paper 🚨
What if LLMs could tell you they’re going to fail before they finish reasoning?
We trained models to predict their own future: whether they’ll succeed and how long it will take. At every token, in real time, with no extra compute.
We used this to develop an adaptive sampling algorithm for test-time compute. 👇🧵
Tim Seyde retweeted
Meet the strongest 3B model on the market.
LFM2-2.6B-Exp is an experimental checkpoint built on LFM2-2.6B using pure reinforcement learning.
> Consistent improvements in instruction following, knowledge, and math benchmarks
> Outperforms other 3B models in these domains
> Its IFBench score surpasses DeepSeek R1-0528, a model 263x larger
Download and play 👉 huggingface.co/LiquidAI/LFM2…
Happy holidays,
The Liquid AI team🎄✨
