Vendor · Language models · Audio models · Decision models · Specialist models
Liquid AI has 9 models in the directory: 5 language models, 1 audio model, 2 decision models and 1 specialist model.
In the directory: 5 language models, 1 audio model, 2 decision models and 1 specialist model. 9 of them with open weights that can run locally.
Every model there ever was. Models with a profile lead to every benchmark, price and source.
| LFM3-1.7B | Oct 27, 2026 | Open weights | n/a | n/a | n/a | 262K | below 1 |
| LFM3-1.7B-Base | Oct 27, 2026 | Open weights | n/a | n/a | n/a | 262K | below 1 |
| LFM3-Omni-2B | Oct 27, 2026 | Open weights | n/a | n/a | n/a | 33K | below 1 |
| LFM3-VL-2B | Oct 27, 2026 | Open weights | n/a | n/a | n/a | 131K | below 1 |
| LFM2 24B A2B | Feb 24, 2026 | Active | n/a | n/a | n/a | 33K | 38 |
| LFM3-Audio-2B | Oct 27, 2026 | Open weights | n/a | below 1 |
| d1-2B | Oct 27, 2026 | Open weights | n/a | below 1 |
| d1-Omni-2B | Oct 27, 2026 | Open weights | n/a | below 1 |
| LFM2.5-VL-3B-DSpark | Sep 24, 2026 | Open weights | n/a | 32 |
Decision model · Oct 27, 2026
d1-2B is a local decision model with 1.92 billion parameters. It reads text, images, and video and answers typed questions in one forward pass without generating answer tokens. Built on LFM3-VL-2B, it supports choice probabilities, extraction, boxes, and estimation intervals.
Decision model · Oct 27, 2026
d1-Omni-2B accepts text, images, video, and audio and returns typed, calibrated answers in one forward pass. It has about 2.17 billion parameters and builds on LFM3-Omni-2B. This supports structured decisions over spoken and visual input.
Language model · Oct 27, 2026
LFM3-1.7B is a dense text model from Liquid AI with 1.73 billion parameters and 262,144 context tokens. Its 24 layers combine short convolutions with local and global attention. Liquid AI post-trained it for local agents, tool calling, structured responses, and task-specific adaptation.
Language model · Oct 27, 2026
LFM3-1.7B-Base is the pre-trained text checkpoint with 1.73 billion parameters, a 128,000-token vocabulary, and 262,144 context tokens. Liquid AI reports 22 trillion total pre-training tokens. The model is intended for substantial fine-tuning.
Audio model · Oct 27, 2026
LFM3-Audio-2B accepts text and 16 kHz audio and returns text, 24 kHz speech, or interleaved text and speech. Its causal audio encoder and LiquidCodec support streaming. The model has about two billion parameters and 32K context tokens.
Language model · Oct 27, 2026
LFM3-Omni-2B combines text, image, video, and audio in one local model. It processes video with its soundtrack, accepts 16 kHz audio, and returns text, 24 kHz speech, or both. The card lists 32K context tokens and a default of 2 fps with 4 to 384 video frames.
Language model · Oct 27, 2026
LFM3-VL-2B accepts text, native-resolution images, and video. The dense model has 1.92 billion parameters, a 128K context window, and an in-house vision encoder. Its card publishes 62 visual benchmarks with thinking enabled and disabled.
Specialist model · Sep 24, 2026
LFM2.5-VL-3B-DSpark is a speculative-decoding draft model. It proposes candidates for the target model LFM2.5-VL-3B, which verifies and corrects them when needed. The draft and target model form one inference system. DSpark accelerates local decoding for the target model, with open weights aimed at runtimes such as SGLang and MLX-VLM.
Language model · Feb 24, 2026
LFM2 24B A2B is the largest model of Liquid AI's LFM2 series, which is designed for on-device use. The MoE model has 24 billion total and 2.3 billion active parameters per token.
| Vendor | Models | Model types | Best language model | Top 10 | Open weights | New in 2026 |
|---|---|---|---|---|---|---|
| Anthropic | 33 | 33 language models | Claude Opus 5.5 (Rank 1 to 3) | 4 | 0 | 13 |
| OpenAI | 63 | 46 language models, 4 image models, 2 video models, 9 audio models, 2 specialist models | GPT-6 Astra (Rank 4 to 8) | 2 | 4 | 27 |
| 66 | 43 language models, 4 image models, 6 video models, 11 audio models, 2 specialist models | Gemini 4 Carbon (Rank 1 to 5) | 2 | 19 | 26 | |
| Meta | 32 | 29 language models, 1 image model, 1 video model, 1 specialist model | Muse Wonder Full (Rank 6 to 9) | 2 | 26 | 7 |
| DeepSeek | 23 | 23 language models | DeepSeek V4.1 Flash 1017 (Rank 13 to 20) | 0 | 23 | 9 |
| Moonshot AI | 12 | 12 language models | Kimi K3.1 Pro (Rank 10 to 13) | 0 | 10 | 7 |
| xAI | 16 | 12 language models, 2 image models, 2 video models | Grok 4.7 (Rank 18 to 24) | 0 | 1 | 12 |
| Liquid AI | 9 | 5 language models, 1 audio model, 2 decision models, 1 specialist model | not ranked yet | 0 | 9 | 9 |
By release date. Grey: retired models.
9 releases since January 2026, by release date.
The directory lists 9 Liquid AI models: 5 language models, 1 audio model, 2 decision models and 1 specialist model. It counts every model Liquid AI ever released, including retired ones. 9 are currently available, the newest being d1-2B, d1-Omni-2B and LFM3-1.7B.
d1-2B, released on Oct 27, 2026. The timeline below shows every release by date.
Yes, 9 of the 9 models have open weights. Each model profile lists how much memory it needs.
Up to 262,144 tokens (≈ 390 A4 pages), which 2 of the current models reach, including LFM3-1.7B. Every model's value is in the model list.
Data as of Nov 1, 2026