Liquid AI
5 models across 2 families · Latest: LFM2.5 230M (2026-06)
Adaptive and efficient AI models
Liquid AI's portfolio covers 5 active models across 2 current families, spanning rag, agents, and long context. Open a model detail page to compare provider routes and sourced benchmarks.
Covers 5 workload areas across 5 active tracked models; last verified 2026-07-10.
Use it for
- Teams evaluating rag, agents, and long context across this lab's releases
- Comparing model families before committing to a flagship
- Migration and pricing follow-ups across 5 tracked models
Do not use it for
- Choosing a hosting provider without opening a model page for price ladders
Active models
5
Current models from this lab, excluding deprecated ones
Active families
2
Current model families from this lab
Open catalog
5 open
0 open source / 5 open weights
Lowest output price
$0.120 /1M
Cheapest tracked output across active models, per 1M tokens
Latest dated release
2026-06-25
LFM2.5 230M
Freshness
2026-07-10
Researched 8d ago
Information
Release cadence
Showing 5 recent dated releases (full timeline below). Latest: LFM2.5 230M (2026-06-25).
Where this lab wins
- RAG: 1 tracked model with ruler / needle retrieval benchmarks.
- Agentic: 4 tracked models with BFCL, tau-bench, and SWE-bench tool-use coverage.
- Long-context: 1 tracked model with context-token or InfiniteBench-class signal.
- Classification: 3 tracked models with MMLU-class moderation/safety coverage.
Flagship quality / price signal
Flagship: Together LFM2-24B (best sourced coding quality-per-dollar in this portfolio).
Quality-per-dollar unavailable for this flagship — benchmark coverage or output token pricing is still missing.
Liquid AI is an American AI research organization founded in 2023. Adaptive and efficient AI models. Liquid AI ships 2 model families totaling 5 models, with the most recent release LFM2.5 230M in 2026-06. Notable families include LFM-2.5 and LFM-2. Use it as a stable reference for lab background, release coverage, and follow-up model pages as they are added. Researchers and evaluators. View official API endpoints, benchmark performance, and coding/agent fit for every Liquid AI model.
About
Liquid AI, an MIT spin-off established in 2023, has rapidly positioned itself at the forefront of generative AI and large language model (LLM) development. Based in Cambridge, Massachusetts, the company's mission is to democratize AI solutions, ensuring they are accessible and controllable by users. Liquid AI's innovative approach transcends traditional transformer-based architectures, marking a significant evolution in the capabilities and efficiency of modern AI systems. Central to Liquid AI's innovation is the creation of Liquid Foundation Models (LFMs), which define a new standard in generative AI technology. These models deliver state-of-the-art performance across various benchmarks, requiring less memory and providing more efficient inference compared to conventional LLMs. Their unique ability to handle diverse data types such as video, audio, text, and time series, empowers broader applications across industries. This versatility is underpinned by their distinctive architecture, which incorporates computational units leveraging principles from dynamical systems, signal processing, and numerical linear algebra, facilitating efficient long-sequence processing. Beyond model development, Liquid AI emphasizes operational and training efficiency. By optimizing their pre- and post-training infrastructure, the company maximizes performance in areas such as multi-step reasoning, long-context recall, and knowledge retention. They distinguish themselves through their "white-box" philosophy, ensuring explainability and transparency in an industry often dominated by opaque "black box" models. Their commitment to open science is evident through active contributions to research and collaborative efforts within the AI community, fostering innovation and accelerating advancements in the field. Liquid AI has already marked significant achievements with the release of language models such as LFM-1B, LFM-3B, and LFM-40B, all of which exhibit exemplary performance within their respective categories. The company has also attracted $46.6 million in seed funding, reflecting robust investor confidence in their pioneering technology and ambitious vision. Collaborations with entities like ITOCHU Techno-Solutions Corporation highlight their expanding global reach and influence. This strategic engagement not only broadens their impact but also underscores the company's role as a trailblazer in advancing AI capabilities.
Featured models
| Model | Released | Context | Input price ($/1M) | Output price ($/1M) | License | Openness |
|---|---|---|---|---|---|---|
| LFM2.5 230M | 2026-06-25 | 32k | - | - | LFM Open License v1.0 | Open weights |
| LFM2.5 8B A1B | 2026-05-28 | 128k | - | - | LFM Open License v1.0 | Open weights |
| LFM2.5 1.2B Instruct | 2026-02-01 | 32k | - | - | LFM Open License v1.0 | Open weights |
Model families
Recent releases
- LFM2.5 230M- 2026-06-25
- LFM2.5 8B A1B- 2026-05-28
- LFM2.5 1.2B Instruct- 2026-02-01
- Together LFM2-24B- 2025-12-01
- LFM2-24B-A2B- 2025-11-01
FAQ
Who founded Liquid AI and when?
Liquid AI was founded in 2023 and is associated with Cambridge, MA, United States.
What models has Liquid AI released?
Liquid AI ships 5 models across 2 families: LFM-2.5 and LFM-2.
Is Liquid AI's technology open source?
All tracked Liquid AI models are open-weight or open-source.
Where is Liquid AI headquartered?
Liquid AI is headquartered in Cambridge, MA, United States.
What is Liquid AI known for?
Adaptive and efficient AI models. Its most prominent tracked family is LFM-2.5.
How can I access Liquid AI's models?
Liquid AI's models are available via Liquid AI and OpenRouter.
Explore related pages
Last reviewed: 2026-07-10. Data sourced from public lab announcements and provider documentation.