Liquid Ai
PulseAugur coverage of Liquid Ai — every cluster mentioning Liquid Ai across labs, papers, and developer communities, ranked by signal.
- developed LFM2.5-24B-A2B 95%
- instance of LFM2.5-2.6B 95%
- developed LFM3 95%
- developed LFM2.5-VL-3B 95%
- developed d1-omni-600M 95%
- instance of d1-3B 95%
- developed d1-3B 95%
- developed LFM2.5-230M 95%
- instance of LFM2.5-230M 95%
- instance of LFM3-VL-2B 95%
- developed LFM3-VL-2B 95%
- instance of LFM2.5-8B-A1B 95%
- 2026-10-28 product_launch Liquid AI released the LFM3 family of small, open-weight, multimodal models for local and on-device use. source
- 2026-10-08 product_launch Liquid AI released the open-source decision-making models d1-3B and d1-omni-600M. source
- 2026-10-07 product_launch Liquid AI released the open-weight multimodal decision models d1-3B and d1-omni-600M. source
- 2026-08-22 product_launch Liquid AI is reportedly developing a new 100 billion parameter model. source
- 2026-08-20 product_launch Liquid AI released DSpark draft models for its LFM2.5 family, offering up to 3.18x faster decoding speeds. source
- 2026-08-20 product_launch Liquid AI released DSpark draft models for its LFM2.5 series, enhancing decoding speed. source
- 2026-07-29 product_launch Liquid AI released LFM2.5-Encoders, a model for fast long-text analysis on CPUs. source
- 2026-07-29 product_launch Liquid AI released two new open-weight bidirectional encoder models, LFM2.5-Encoder-230M and LFM2.5-Encoder-350M. source
- 2026-06-19 product_launch Liquid AI released two new retrieval models, LFM2.5-Embedding-350M and LFM2.5-ColBERT-350M, designed for efficient multilingual search on edge devices. source
- 2026-06-19 product_launch Liquid AI released two new retrieval models for multilingual search. source
- 2026-05-29 product_launch Liquid AI released its LFM2.5-8B-A1B model, featuring an expanded context window and increased training data. source
- 2026-05-26 product_launch Liquid AI launched two new Japanese-capable AI models, LFM2.5-1.2B-JP-202606 and LFM2.5-Audio-1.5B-JP. source
20 day(s) with sentiment data
Liquid AI's d1 models to see rapid adoption for structured decision tasks
Liquid AI's recent releases of d1-3B and d1-omni-600M, specifically designed for structured decision-making with calibrated probabilities and zero output tokens, suggest a strong market fit. Their focus on efficiency, predictable latency, and cost-effectiveness compared to general LLMs positions them well for tasks like classification, routing, and scoring, which are common in enterprise applications. Early user reports on consumer hardware performance further support the potential for widespread adoption.
Liquid AI's open-source strategy is driving performance and accessibility
The release of d1-3B and d1-omni-600M as open-source models on Hugging Face, with support for local deployment via llama.cpp, indicates a strategic push towards accessibility and community engagement. This aligns with the broader trend of open-weight models gaining traction, as evidenced by the success of models like Darwin-180B-RSI. Making these specialized decision models readily available lowers the barrier to entry for developers and businesses seeking efficient decision-making solutions.
Liquid AI to release enterprise-focused solutions leveraging d1 models
Given Liquid AI's emphasis on predictable latency, cost-effectiveness, and calibrated probabilities for structured decisions, it's plausible they will soon offer enterprise-grade solutions or APIs built around their d1 model family. This would cater to businesses looking to integrate reliable, efficient decision-making capabilities into their workflows without the overhead of general-purpose LLMs. The optimization for NVIDIA hardware also hints at a potential focus on scalable enterprise deployments.
Liquid AI's LFM2.5-2.6B model to see widespread adoption in on-device agent applications within 6 months
The release of LFM2.5-2.6B, with its strong emphasis on agentic capabilities, compact size, and ability to run on-device with a large context window, positions it well for use in applications requiring local AI processing. The partnership with MacPaw for macOS integration further suggests a strategic push towards broader adoption in consumer devices. We expect to see this model integrated into various agentic applications targeting mobile and desktop within the next six months.
Liquid AI to release a fine-tuned version of LFM2.5-2.6B optimized for coding tasks within 3 months
While current reports indicate LFM2.5-2.6B is not recommended for complex coding, its compact size and agentic focus make it a strong candidate for specialized fine-tuning. Given the general demand for capable coding assistants, Liquid AI may prioritize developing a version of this model specifically for code generation or assistance, leveraging its efficient architecture. We hypothesize such a release within the next three months.
-
Liquid AI's LFM3-1.7B model fits in 1GB, claims benchmark win over Qwen3.5-9B
Liquid AI has released LFM3-1.7B, a new 1.73 billion parameter text model that can fit into 1 GB of memory and process up to 256K tokens of context. While Liquid AI's own benchmarks suggest LFM3-1.7B outperforms Qwen3.5…
-
Liquid AI unveils efficient 1GB LFM3 model family for local agents
Liquid AI has released LFM3, a new family of open-weight models designed for efficiency and local deployment. The flagship LFM3-1.7B model, weighing approximately 1 GB, boasts a 256K-token context window and a unique tr…
-
Ollama enables local text-to-speech with LFM3-audio model
Ollama has updated to version 0.40.13, enabling local text-to-speech (TTS) capabilities on Linux and Windows. This update integrates Liquid AI's lfm3-audio model, a 2B parameter model that can generate high-quality spee…
-
Build a local RAG API with FastAPI, ChromaDB, and LFM3 on Ollama
This article details how to build a fully local Retrieval-Augmented Generation (RAG) API using FastAPI, ChromaDB, and the LFM3 model from Liquid AI, all running through Ollama. The author emphasizes a simplified approac…
-
Liquid AI's LFM3 Omni 2B model runs in-browser via WebGPU
Liquid AI has unveiled LFM3 Omni 2B, a new omni-model capable of running within a web browser using WebGPU. A demonstration video showcases the model's performance in this environment.
-
Liquid AI releases LFM3-Omni-2B multimodal model with voice output
Liquid AI has released LFM3-Omni-2B, a new 2B parameter model capable of processing text, images, video, and audio, and responding with text or 24 kHz speech. While Ollama and LM Studio can load the model, only llama.cp…
-
Liquid AI releases LFM3 family of six open-weight multimodal models
Liquid AI has released its LFM3 family, comprising six multimodal models with a total of 1.7 billion parameters. These models are open-weight, and have garnered significant attention on Hugging Face, accumulating 559 li…
-
Ollama v0.40.13 adds support for Liquid AI's LFM3 models on Linux and Windows
Ollama has released version 0.40.13, which now supports Liquid AI's LFM3 models. These LFM3 models, including capabilities for text-to-speech via lfm3-audio and the lfm3-omni model, are now operational on both Linux and…
-
Liquid AI releases LFM3-VL-2B with improved OCR, similar size
Liquid AI has released LFM3-VL-2B, a multimodal model that offers improved OCR capabilities compared to its predecessor, LFM2-VL-1.6B, while maintaining a similar download size. The new model features a significantly re…
-
Liquid AI releases LFM3 family of 2B multimodal open-weight models
Liquid AI has introduced LFM3, a new series of six open-weight models, each approximately 2 billion parameters in size. These models share a common architecture and are designed to handle diverse tasks including text, v…
-
Liquid AI releases LFM3 family of small, open-weight, multimodal models
Liquid AI has released LFM3, a new family of small, open-weight models designed for local and on-device deployment. These models are capable of handling multiple modalities, offering an alternative to large, cloud-based…
-
Google unveils Genie 4, Liquid AI launches LFM3, Anthropic faces scrutiny
Google has released Genie 4, an advanced world model capable of generating interactive, photorealistic environments with synchronized audio and real-time editing, doubling the duration of its predecessor. Liquid AI has …
-
Ollama adds support for Liquid AI's LFM3 models on Linux and Windows
Ollama has released version v0.40.13, which now supports Liquid AI's LFM3 models on Linux and Windows. This update expands the availability of LFM3, a family of small models designed for local agents and tool calling, t…
-
Liquid AI releases 6 open-weight 2B multimodal models, LFM3
Liquid AI has launched LFM3, a new family of six open-weight, 2B-class multimodal models. These models, built on a shared backbone, are designed for text, vision, audio, and decision-making tasks, with some capable of h…
-
Liquid Ai releases LFM3 open models for local small business use
Liquid Ai has released a new family of open-weight models called LFM3, designed for small businesses and local deployment. The LFM3-1.7B model, which fits into approximately 1 GB of memory, is particularly suited for ta…
-
Liquid Ai unveils LFM3 multimodal models for specialized intelligence · 4 sources tracked
Liquid Foundation Models has released LFM3, a new family of multimodal models designed for specialized intelligence. Built on a hybrid architecture and trained on 22T tokens, LFM3 offers improved expressivity and effici…
-
AI shifts to specialized decision models, challenging generative LLMs
The AI landscape is shifting towards specialized decision models, moving away from general-purpose generative LLMs for tasks like routing or validation. These decision models offer typed probabilities for predefined ans…
-
Liquid AI releases LFM2.5-24B-A2B for efficient agent tasks
Liquid AI has released LFM2.5-24B-A2B, a large open-weight model designed for agentic tasks. This model utilizes a mixture-of-experts approach, activating approximately 2 billion parameters per token out of its total 24…
-
LLM pipeline research questions cost-effectiveness; Nemotron 3.5 and LFM2.5 deployment details emerge
A research paper explores the trade-offs in multi-agent LLM pipelines, finding that while protocols like Open Floor Protocol add minimal overhead, decomposing tasks across different models does not improve accuracy or s…
-
Liquid AI launches RIPL engine for sub-50ms LFM inference
Liquid AI has introduced RIPL, a new real-time inference engine designed to serve its Liquid Foundation Models (LFMs) with very low latency. The engine aims to enable the use of generative models in applications that pr…