Liquid
About Liquid
Liquid AI is an efficiency-first foundation model company. We build highly capable, compute-optimized models that bring intelligence to any device and medium of choice.
Our mission is to build efficient general-purpose AI at every scale. This means Liquid products are highly capable, compute and cost optimized, and ready to bring intelligence on any device and medium of choice.
Demand for capable and practical intelligence compounds exponentially . Yet we do not have the energy and capex infrastructure to serve it. Cloud-only serving is intractable on cost, power, and latency . With the rise of open source, enterprises have started realizing that they can own intelligence instead of renting it . We think enterprises should own their data, intelligence, and decision power. An untapped opportunity to complement cloud-based intelligence is the physical world : nearly 2.2 billion devices that can host a foundation model ship every year, across phones, PCs, wearables, and cars alone, before counting robots and machines. Serving powerful AI on these surfaces creates a large market opportunity, enabling us to deliver intelligence efficiently to all .
To make intelligence ubiquitous, we must co-design models with hardware , from first principles. Hardware and use cases set hard limits on memory, battery, and latency , and devices are fragmented across CPUs, GPUs, and NPUs from many chip makers. Meeting these limits becomes even harder when frontier quality is required for models that must fit the constraints of the processor. Tailoring the model to each chip makes every device present as one substrate . Therefore, foundation model training technology must have the hardware in the loop . This way, the fragmentation of on-device processors and low-level software can be solved at once and agents can flourish on top of the local intelligence layer .
Website ↗GitHub ↗Hugging Face ↗Documentation ↗
Source ↗ · 2026-11-12T02:17:07.490660271Z8 indexed models
8 matching models
liquid/lfm-2.5-2.6b-20260811
LFM2.5-2.6B is a compact reasoning model from Liquid AI. It is suited for agent workflows, data extraction, RAG, and long-context processing. Liquid advises against using it for agentic coding or...
- Context window
- 128,000
- Input modalities
- text
- Output modalities
- text
Benchmarks
liquid/d1-20260930
d1 is Liquid AI's structured decision model, served as a System One endpoint. Send a state along with typed questions, and it returns a choice, a score, or a yes/no answer, each with a probability taken directly from the model rather than written out as text. It uses the same /v1
- Context window
- 65,536
- Input modalities
- text, image
- Output modalities
- decisions
- Capabilities
- decisions, multimodal, promptCaching, visionInput
Inference prices
- Published price : $0.02 / cache_read_tokens_per_million · openRouter ↗
- Published price : $0.02 / input_tokens_per_million · openRouter ↗
liquid/d1.1-20261027
d1.1 is Liquid AI's structured decision model, served as a System One endpoint and the successor to d1. Send a state along with typed questions, and it returns a choice, a score, or a yes/no answer, each with a probability taken directly from the model rather than written out as
- Context window
- 65,536
- Input modalities
- text, image
- Output modalities
- decisions
- Capabilities
- decisions, multimodal, promptCaching, visionInput
Inference prices
- Published price : $0.02 / cache_read_tokens_per_million · openRouter ↗
- Published price : $0.02 / input_tokens_per_million · openRouter ↗
liquid/d1-ultra-20261015
d1-ultra is Liquid AI's largest decision model, served as a System One endpoint. Send a state of text and images along with typed questions, and it returns a choice, a score, or a yes/no answer, each with a probability taken directly from the model rather than written out as text
- Context window
- 32,768
- Input modalities
- text, image
- Output modalities
- decisions
- Capabilities
- decision, decisions, multimodal, promptCaching, visionInput
Inference prices
- Published price : $0.04 / cache_read_tokens_per_million · openRouter ↗
- Published price : $0.04 / input_tokens_per_million · openRouter ↗
liquid/lfm-2.5-24b-a2b-20261012
LFM2.5-24B-A2B is a mixture-of-experts reasoning model from Liquid AI with 24B total and 2.45B active parameters. It always reasons before answering and is built for agents that call tools. Liquid recommends pairing it with retrieval or web search for knowledge-heavy tasks.
- Context window
- 128,000
- Input modalities
- text
- Output modalities
- text
- Capabilities
- promptCaching, reasoning, structuredOutputs, toolCalling
Inference prices
- Published price : $0.02 / cache_read_tokens_per_million · openRouter ↗
- Published price : $0.05 / input_tokens_per_million · openRouter ↗
- Published price : $0.70 / output_tokens_per_million · openRouter ↗
liquid/lfm-2.5-embedding-350m-20260818
LFM2.5-Embedding-350M is a text embedding model from Liquid AI. It produces 1,024-dimensional embeddings for retrieval and semantic search. Successful OpenRouter requests and embeddings may be retained and used to train...
- Context window
- 512
- Input modalities
- text
- Output modalities
- embeddings
Benchmarks
liquid/lfm-3-1.7b-20261106
A compact 1.7B-parameter text model with a 256K-token context, post-trained for agents, tool calling, and structured outputs. It reasons before it answers. Prompts and outputs may be retained and used to train Liquid models.
- Context window
- 262,144
- Input modalities
- text
- Output modalities
- text
- Capabilities
- promptCaching, reasoning, structuredOutputs, toolCalling
Inference prices
- Published price : $0.02 / cache_read_tokens_per_million · openRouter ↗
- Published price : $0.08 / input_tokens_per_million · openRouter ↗
- Published price : $0.20 / output_tokens_per_million · openRouter ↗
vercelAIGatewayCatalog/liquid/d1
Liquid d1 is a decision model for classification, routing, and scoring. It evaluates shared state against typed Choice, Score, and yes/no questions and returns structured answers with calibrated probabilities through the TypeSafe-compatible System One API.
- Context window
- 65,536
- Input modalities
- text, image
- Output modalities
- text
- Capabilities
- decision, multimodal, visionInput
Inference prices