LFM2.5-8B-A1B
The memory it needs, what you may do with it, and the hardware it suits.

LFM2.5-8B-A1B is a language model for chat, drafting and question answering, published by Liquid AI on Hugging Face on 28 May 2026. It has about 8.5 billion parameters and a configured context window of 128,000 tokens. It is a mixture-of-experts model: 32 experts, 4 used for each token, so only about 1B parameters are active at a time.

Its card lists the licence as a custom licence (lfm1.0). In short, the licence sets its own terms; read it before any commercial use.

Hugging Face counted about 26,435 downloads of it in the 30 days before we checked (7 Oct 2026), and 784 people have liked it there.

The card lists 10 languages: en, ar, zh, fr, de, ja, ko, es, pt, it (ISO codes as given on the card).

It is built on LiquidAI/LFM2.5-8B-A1B-Base, according to its card.

We size it the way you would size any model: the weights (parameters × bits ÷ 8), the KV cache for 8,192 tokens of context, and roughly 1.5 GB for the runtime itself. Real use varies with settings.

Each token of context adds about 0.01 MB of KV cache at 16-bit, so 32,768 tokens would add about 0.40 GB. Only 6 of its 24 layers use standard attention; the others keep a small fixed state, so the cache is smaller than in a classic transformer.

Arithmetic against the launch specification, not a test: about three quarters of ANVE One’s 64 GB of unified memory for the model (as AMD’s Variable Graphics Memory allows), and the 24 GB on the graphics card in ANVE Office and ANVE Office Pro, which keep the rest of the stack in 64 GB and 128 GB of system memory. ANVE’s tested default is Qwen 3.8 27B; this model is not yet validated on ANVE.

This size suits quick drafting, classification, extraction and chat over short documents on modest hardware. ANVE pairs a fast 4B-class model with its main model for quick jobs.

• LFM2-8B-A1B: a variant with the same size and memory needs (same parameter count, so the memory figures above apply).

• Mistral-NeMo-Minitron-8B-Instruct (8.4B parameters, about 6.2 GB at 4-bit before context; custom licence: check before commercial use)
• codegemma-7b-it (8.5B parameters, about 6.3 GB at 4-bit before context; commercial use allowed, with conditions)
• Qwen3-8B (8.2B parameters, about 6.1 GB at 4-bit before context; commercial use allowed)

No sealed ANVE unit exists; every image of the appliance on this site is a labelled render. The software runs today on a development machine. No customers yet, no certifications, no delivery date, and the final price is not announced.

Reservations are ₹999 in India or US$12 elsewhere, refundable in full until you confirm an order in writing (terms). Offices that cannot use cloud AI can apply to be one of ten design partners.