// Open-model VRAM table · family
Liquid LFM: local deployment & VRAM
26 open-weight models from Liquid AI (138 official repos incl. quantised builds). At 4-bit they range from 0.8 GB (LFM2.5-Encoder-350M-Diffusion) to 16.7 GB (LFM2.5-24B-A2B).
official orgs: LiquidAIlatest 2026-10-12memory = estimate
01Text LLM (16)
| Model | Params | Context | BF16 | 4-bit | Smallest GPU (4-bit) | Mac (4-bit) | Licence | Released |
|---|---|---|---|---|---|---|---|---|
| LFM2.5-24B-A2B | 24B / 2B | 128K | 53.4 GB | 16.7 GB | RTX 4090 24GB | Mac 36GB | Custom licence – read it | 2026-10-12 |
| LFM2.5-2.6B | 2.7B | 128K | 6.6 GB | 2.4 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2026-07-28 |
| LFM2.5-Encoder-350M-Diffusion ↗ | 354M | 125K | 1.4 GB | 0.8 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2026-07-10 |
| LFM2.5-230M | 230M | 125K | 1.1 GB | 0.8 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2026-06-24 |
| LFM2.5-8B-A1B | 8.47B / 1B | 125K | 19.2 GB | 6.2 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2026-05-28 |
| LFM2.5-1.2B-JP-202606 | 1.17B | 125K | 3.2 GB | 1.4 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2026-05-26 |
| LFM2.5-350M | 354M | 125K | 1.4 GB | 0.8 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2026-03-31 |
| LFM2-24B-A2B | 23.8B / 2B | 125K | 53.1 GB | 16.6 GB | RTX 4090 24GB | Mac 36GB | Custom licence – read it | 2026-02-24 |
| LFM2.5-1.2B-Thinking | 1.17B | 125K | 3.2 GB | 1.4 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2026-01-20 |
| LFM2.5-1.2B-Instruct | 1.17B | 125K | 3.2 GB | 1.4 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2026-01-06 |
| LFM2.5-1.2B-JP | 1.17B | 125K | 3.2 GB | 1.4 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2026-01-04 |
| LFM2-8B-A1B | 8.34B / 1B | 125K | 18.9 GB | 6.2 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2025-10-07 |
| LFM2-2.6B | 2.57B | 125K | 6.3 GB | 2.3 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2025-09-22 |
| LFM2-350M | 354M | 125K | 1.5 GB | 1.0 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2025-07-10 |
| LFM2-700M | 742M | 125K | 2.4 GB | 1.3 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2025-07-10 |
| LFM2-1.2B | 1.17B | 125K | 3.3 GB | 1.6 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2025-07-10 |
02Vision-language (6)
| Model | Params | Context | BF16 | 4-bit | Smallest GPU (4-bit) | Mac (4-bit) | Licence | Released |
|---|---|---|---|---|---|---|---|---|
| LFM2.5-VL-3B | 3.12B | 32K | 7.5 GB | 2.7 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2026-08-11 |
| LFM2.5-VL-450M | 449M | 125K | 1.6 GB | 0.9 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2026-04-08 |
| LFM2.5-VL-1.6B | 1.6B | 125K | 4.1 GB | 1.7 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2026-01-05 |
| LFM2-VL-3B | 3B | 125K | 7.2 GB | 2.6 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2025-10-22 |
| LFM2-VL-450M | 451M | 125K | 1.6 GB | 0.9 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2025-08-12 |
| LFM2-VL-1.6B | 1.58B | 125K | 4.1 GB | 1.7 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2025-08-12 |
03Embedding (3)
| Model | Params | BF16 | 4-bit | Smallest GPU (4-bit) | Mac (4-bit) | Licence | Released |
|---|---|---|---|---|---|---|---|
| LFM2.5-ColBERT-350M | 353M | 1.3 GB | 0.7 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2026-05-20 |
| LFM2.5-Embedding-350M | 354M | 1.3 GB | 0.7 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2026-05-05 |
| LFM2-ColBERT-350M | 353M | 1.3 GB | 0.7 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2025-10-28 |
04Text-to-speech / audio (1)
| Model | Params | BF16 | 4-bit | Smallest GPU (4-bit) | Mac (4-bit) | Licence | Released |
|---|---|---|---|---|---|---|---|
| LFM2.5-Audio-1.5B | 1.47B | 3.7 GB | 1.5 GB | 8GB GPU | Mac 16GB | Custom licence – read it | 2025-12-18 |
05Other families
Qwen · IBM Granite · NVIDIA Nemotron · Gemma · Mistral / Magistral / Devstral · GLM / CogVideoX (Z.ai) · DeepSeek · Ai2 OLMo / Molmo · Ant Ling / Ring / Ming · MiniCPM / VoxCPM · Tencent Hunyuan / Hy · Phi · Llama · Meituan LongCat · Jina embeddings & rerankers · Wan · Kimi · Cohere Command / Aya · Xiaomi MiMo · FLUX · Microsoft VibeVoice / Fara / Harrier · InternLM / Intern-S · StepFun Step · MiniMax · Baidu ERNIE · NVIDIA Parakeet / Canary speech · Stable Diffusion / Stable Audio · Yi (01.AI) · ByteDance Seed · SmolLM / SmolVLM · BGE embeddings & rerankers · LTX Video · FunAudioLLM speech · gpt-oss · poolside Laguna · Whisper · Baichuan · Thinking Machines Inkling · xAI Grok · rednote dots
Totals include 10% + 0.5GB overhead and one 8K-token KV cache for text models. Estimates only. Method →