Model library / LiquidAI
LFM2.5-24B-A2B-Base is a 24 billion-parameter mixture-of-experts model from LiquidAI aimed at on-device deployment and heavy fine-tuning.
Mixture of experts, count not published
Context length in tokens, as published
Weights at the released precision
Licence declared on the repository
Overview
LFM2.5-24B-A2B-Base is a 24 billion-parameter mixture-of-experts language model released by LiquidAI. It contains 40 layers, 64 experts per MoE block with top-4 routing, and 2.45 billion active parameters. The model was trained on 36 trillion tokens, supports a 128 k vocabulary and 17 languages, and is intended for on-device use and specialised fine-tuning such as language-specific or domain-specific assistants.
In practice the model occupies a large memory footprint due to its 24 billion total parameters, but only 2.45 billion are active at inference. It accepts up to 131 072 tokens of context, uses bfloat16 weight format, and is released under an "other" licence. The repository provides inference support through multiple frameworks and is compatible with Transformers ≥ 5.0.0.
Specification
| Repository | LiquidAI/LFM2.5-24B-A2B-Base |
| Publisher | LiquidAI |
| Published | 9 October 2026 |
| Architecture | Lfm2MoeForCausalLM |
| Model type | lfm2_moe |
| Parameters | 23,971,585,280 |
| Layers | 40 |
| Hidden size | 2,048 |
| Attention heads | 32, 8 key/value heads |
| Context length | 131,072 tokens |
| Vocabulary | 128,000 tokens |
| Weight format | bfloat16 |
| Licence | other |
Memory
Parameter count multiplied by the bytes each format uses. The key-value cache comes on top.
| Precision | Weights | Notes |
|---|---|---|
| bf16 | 44.7 GiB | Full precision as released |
| fp8 | 22.3 GiB | 8-bit, near-lossless on most models |
| int4 | 11.2 GiB | 4-bit, smallest footprint |
The cache costs 80 KiB per token at 16-bit, so the full 131,072-token context of one request needs about 10.0 GiB. Concurrency multiplies that number, not the weights.
Hardware
At the released precision, with 90% of the card given to the server and 2.5 GiB kept for the runtime. The context column is what is left for the cache on that setup.
| GPU | Cards needed | Free for cache | Context it holds |
|---|---|---|---|
| RTX 4090 24GB | 4 | 31.7 GiB | the full 131,072 tokens |
| RTX 5090 32GB | 2 | 7.9 GiB | about 104,194 tokens |
| L40S 48GB | 2 | 36.7 GiB | the full 131,072 tokens |
| A100 80GB | 1 | 24.8 GiB | the full 131,072 tokens |
| H100 80GB | 1 | 24.8 GiB | the full 131,072 tokens |
| RTX PRO 6000 96GB | 1 | 39.2 GiB | the full 131,072 tokens |
| H200 141GB | 1 | 79.7 GiB | the full 131,072 tokens |
Serving
A vLLM launch line for the shape above. Check the model card for a runtime the publisher recommends.
vllm serve LiquidAI/LFM2.5-24B-A2B-Base \ --max-model-len 131072 \ --gpu-memory-utilization 0.90
Questions
The weights take about 45 GiB at the released precision. That fits on one A100 80GB, which leaves roughly 25 GiB for the key-value cache.
The repository declares other. Read the licence text before commercial use: the name alone does not tell you what is allowed.
The configuration allows 131,072 tokens. Whether the whole window is usable depends on the memory left for the cache, which the hardware table works out per card.
This page is a reference, and Aiqre does not serve this model today. We deploy open-weight models on dedicated EU hardware on request, so if you want an OpenAI-compatible endpoint for it, email [email protected] with the model name and roughly what volume you expect.
Also in the library
LFM2.5-24B-A2B is a 24 billion-parameter MoE reasoning model with 2.45 billion active parameters for on-device agentic tasks.
LFM2.5-2.6B is a 2.7B open-weight model published by LiquidAI on Hugging Face. It uses the Lfm2ForCausalLM architecture with 30 l…
Hy-Memory-30B-A3B is a Tencent-developed long-term-memory writer model designed to replace LLMs in Mem0, MemOS and Hy-Memory fram…
Granite Switch 4.1 30B is a modular LLM combining a 30-billion base, decision and guardian adapters, and an English speech-to-tex…
Hosting
Tell us the model and roughly what volume you expect. We reply with a price and an OpenAI-compatible endpoint, with your prompts never stored.