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

LFM2.5 DSpark

By Liquid AI · United States

Updated 2026-08-28

chat general small
Parameters
7B
License
LFM Open License v1.0
Context
32k
VRAM (Q4)
4.1 GB

Overview

LFM2.5 DSpark is a dense 7B Liquid Foundation Model variant from Liquid AI, tuned for general-purpose chat on edge and CPU hardware with a 32K context window and roughly 4.1GB VRAM at Q4.

When to pick this model

  • General chat on a modest 6-8GB GPU
  • CPU or edge inference with limited memory
  • Local assistants that don't need long-context or code specialization
  • Lightweight alternative when Ollama distribution isn't required

VRAM requirements by quantization

VRAM REQUIRED (GB)812Q4_K_M4.1 GBQ5_K_M5 GBQ8_07 GBFP1614 GB
QuantizationVRAM required
Q4_K_M (recommended)4.1 GB
Q5_K_M5 GB
Q8_07 GB
FP16 (no quantization)14 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 DSpark fits an 8 GB consumer card at Q4_K_M (4.1 GB). Stepping up to Q8_0 nearly doubles the footprint to 7 GB, and unquantized FP16 weights take 14 GB — plan your GPU around the Q4 or Q5 figure unless you specifically need the higher fidelity.

Without a GPU, LFM2.5 DSpark needs roughly 9 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 32 tokens/sec on entry-level GPUs, on the order of 50 tokens/sec on a mid-range card, and up to 75 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 DSpark 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 DSpark
8 GBRTX 5070 Laptop, RTX 5060, RTX 5060 Ti 8GBQ8_0 (7 GB used)
12 GBRTX 5070, RTX 5070 Ti Laptop, RTX 4080 LaptopQ8_0 (7 GB used)
16 GBRTX 5080, RTX 4080 Super, Radeon RX 9070 XTFP16 (14 GB used)
24 GBRTX 4090, Radeon RX 7900 XTX, RTX 5090 LaptopFP16 (14 GB used)
32 GBRTX 5090FP16 (14 GB used)

Which GPU should you buy to run LFM2.5 DSpark?

To run LFM2.5 DSpark locally at Q4, you need ~4.1 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

  • Dense 7B fits on a 6-8GB GPU at Q4
  • Optimized for edge/CPU inference
  • Versatile general-purpose chat
  • Small memory footprint

Limitations

  • Weights are gated on Hugging Face (requires acceptance)
  • No Ollama tag; install via Hugging Face
  • 32K context is shorter than competing models

Typical workloads

In our catalog grid, LFM2.5 DSpark is filed under General Chat, CPU Inference, Edge/Mobile — the use cases where its size/quality trade-off makes the most sense. Its tags translate to concrete workloads: everyday chat, drafting and summarization.

The 32k-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: Dense Liquid Foundation Model · 7B parameters · 32K context

Training: DSpark variant of Liquid AI's LFM2.5 family, optimized for edge/CPU inference. LFM Open License v1.0.

Verdict

A lightweight 7B general-chat variant for edge and CPU setups with modest GPU memory.

Quick start

# HuggingFace : LiquidAI/LFM2.5-DSpark

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 DSpark need?

At the recommended Q4_K_M quantization, LFM2.5 DSpark needs about 4.1 GB of VRAM. Q8_0 takes 7 GB, and unquantized FP16 weights take 14 GB.

Can LFM2.5 DSpark run without a GPU?

Yes — with roughly 9 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 DSpark support?

LFM2.5 DSpark supports a 32k-token context window (32,768 tokens).

Can I use LFM2.5 DSpark commercially?

LFM2.5 DSpark 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 DSpark on consumer hardware?

Our compatibility engine estimates on the order of 50 tokens/sec on a mid-range GPU and up to 75 tokens/sec on high-end cards, assuming the quantization fully fits in VRAM.

Which quantization of LFM2.5 DSpark should I download first?

Start with Q4_K_M (4.1 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 Q8_0.

Tools

Is LFM2.5 DSpark the right pick for you?

Compute self-hosted ROI → Back to catalog