BestLLMfor Your hardware. Your LLM. Your call.
The Local Copilot Kit APIOpen data Find my LLM
Model fiche

LFM2.5 2.6B

By Liquid AI · United States

Updated 2026-08-28

chat general small multilingual
Parameters
2.6B
License
LFM Open License v1.0
Context
125k
VRAM (Q4)
1.5 GB
Released
2026-07-28

Overview

LFM2.5 2.6B is Liquid AI's ultra-compact dense model with a 128K context window and multilingual support, running in just 1.5GB VRAM at Q4 for CPU and edge use.

When to pick this model

  • Multilingual chat on extremely resource-constrained devices
  • Mobile or embedded deployments needing long-context handling in minimal memory
  • CPU-only inference where even 7B models are too heavy
  • Prototyping edge AI features with a long context budget

VRAM requirements by quantization

VRAM REQUIRED (GB)Q4_K_M1.5 GBQ5_K_M1.8 GBQ8_02.8 GBFP165 GB
QuantizationVRAM required
Q4_K_M (recommended)1.5 GB
Q5_K_M1.8 GB
Q8_02.8 GB
FP16 (no quantization)5 GB

VRAM figures include model weights plus a typical 8k KV cache and ~600 MB runtime overhead (Ollama / llama.cpp baseline). Add headroom for higher context lengths.

In practice, LFM2.5 2.6B fits an 8 GB consumer card at Q4_K_M (1.5 GB). Stepping up to Q8_0 nearly doubles the footprint to 2.8 GB, and unquantized FP16 weights take 5 GB — plan your GPU around the Q4 or Q5 figure unless you specifically need the higher fidelity.

Without a GPU, LFM2.5 2.6B needs roughly 3.4 GB of system RAM to run on CPU via llama.cpp or Ollama — workable for background jobs, but far slower than GPU inference. Throughput estimates from our compatibility engine: around 50 tokens/sec on entry-level GPUs, on the order of 85 tokens/sec on a mid-range card, and up to 130 tokens/sec on high-end hardware — assuming the chosen quantization fully fits in VRAM.

What hardware do you need

The table below matches LFM2.5 2.6B to common GPU memory tiers, using the highest-fidelity quantization that fully fits each card class. Spilling layers to system RAM works but costs most of the speed, so size your card to the quantization you actually want to run.

GPU memoryExample cardsBest fit for LFM2.5 2.6B
8 GBRTX 5070 Laptop, RTX 5060, RTX 5060 Ti 8GBFP16 (5 GB used)
12 GBRTX 5070, RTX 5070 Ti Laptop, RTX 4080 LaptopFP16 (5 GB used)
16 GBRTX 5080, RTX 4080 Super, Radeon RX 9070 XTFP16 (5 GB used)
24 GBRTX 4090, Radeon RX 7900 XTX, RTX 5090 LaptopFP16 (5 GB used)
32 GBRTX 5090FP16 (5 GB used)

Which GPU should you buy to run LFM2.5 2.6B?

To run LFM2.5 2.6B locally at Q4, you need ~1.5 GB of VRAM. The best value for this is a RTX 5060 (8 GB VRAM).

Check RTX 5060 price on Amazon →Check RTX 5060 price on Newegg →

As an Amazon Associate, BestLLMfor earns from qualifying purchases, at no extra cost to you. It does not influence our independent rankings.

Strengths

  • Ultra-compact: 1.5GB VRAM at Q4
  • Native 128K context
  • Multilingual support
  • Optimized for CPU/edge inference

Limitations

  • LFM Open License v1.0 — review terms for commercial use
  • No Ollama tag; install via Hugging Face
  • Small model size limits code and complex-reasoning performance

Typical workloads

In our catalog grid, LFM2.5 2.6B is filed under Multilingual Chat, CPU Inference, Edge/Mobile — the use cases where its size/quality trade-off makes the most sense. Its tags translate to concrete workloads: multilingual workloads.

The 125k-token context window covers long chats and mid-sized documents, though very large retrieval workloads will need chunking. It ships under the LFM Open License v1.0 license — commercial use is generally possible but read the specific terms before embedding it in a product.

Architecture & training

Architecture: Liquid Foundation Model · dense · 2.6B parameters · 128K context

Training: Liquid AI: LFM architecture (an alternative to classic Transformers) optimized for memory efficiency and CPU/edge inference.

Verdict

The smallest LFM2.5 variant — long context and multilingual support in under 2GB of VRAM.

Quick start

# HuggingFace : LiquidAI/LFM2.5-2.6B

Or use the open-source MCP server to query this model from Claude Desktop, Cursor, or any MCP-compatible client.

Similar models worth comparing

Frequently asked questions

How much VRAM does LFM2.5 2.6B need?

At the recommended Q4_K_M quantization, LFM2.5 2.6B needs about 1.5 GB of VRAM. Q8_0 takes 2.8 GB, and unquantized FP16 weights take 5 GB.

Can LFM2.5 2.6B run without a GPU?

Yes — with roughly 3.4 GB of system RAM it runs CPU-only through llama.cpp or Ollama. Expect a fraction of GPU speed, which is fine for background or batch jobs but slow for interactive chat.

What context window does LFM2.5 2.6B support?

LFM2.5 2.6B supports a 125k-token context window (128,000 tokens).

Can I use LFM2.5 2.6B commercially?

LFM2.5 2.6B ships under the LFM Open License v1.0 license. Commercial use is generally permitted subject to its terms — review the license text before shipping a product.

How fast is LFM2.5 2.6B on consumer hardware?

Our compatibility engine estimates on the order of 85 tokens/sec on a mid-range GPU and up to 130 tokens/sec on high-end cards, assuming the quantization fully fits in VRAM.

Which quantization of LFM2.5 2.6B should I download first?

Start with Q4_K_M (1.5 GB) — the standard size/quality sweet spot. Step up to Q5_K_M or Q8_0 only if you have VRAM headroom. It already fits an 8 GB card at FP16.

Tools

Is LFM2.5 2.6B the right pick for you?

Compute self-hosted ROI → Back to catalog