Try LFM • Docs • Discord

LFM2.5-24B-A2B-DSpark-GGUF

GGUF version of LiquidAI/LFM2.5-24B-A2B-DSpark for llama.cpp.

This is a standalone draft sidecar: it carries only the drafter (5 attention layers, rank-256 Markov head, confidence head, block size 9). Token embeddings and the LM head are shared from the target model at load time, so it must be paired with a LFM2.5-24B-A2B-GGUF target file.

Find more information about LFM2.5-DSpark in our blog post.

📦 Files

Choose the target quantization from LFM2.5-24B-A2B-GGUF. Target quantization affects acceptance, speed, and output quality.

🏃 How to run (llama.cpp)

llama-server -hf LiquidAI/LFM2.5-24B-A2B-GGUF:F16 \
  -hfd LiquidAI/LFM2.5-24B-A2B-DSpark-GGUF:F16 \
  --spec-type draft-dspark --spec-draft-n-max 9 --spec-draft-n-min 0 \
  -ngl 99 -ngld 99 -fa on

The block size is read from the sidecar metadata, and --spec-draft-n-max is clamped to it. The target verifies proposed tokens. Floating-point effects can produce different text from target-only decoding, including under greedy sampling. Per-response timings report draft_n and draft_n_accepted.

Other models in the LFM2.5-DSpark GGUF family:

📊 Acceptance and benchmarks

See LiquidAI/LFM2.5-24B-A2B-DSpark for model details and SGLang FP8 benchmarks on NVIDIA B300. Those results are not llama.cpp GGUF measurements; GGUF speedup depends on target quantization, hardware, and workload. See LiquidAI/LFM2.5-24B-A2B for target quality benchmarks.

📬 Contact

• Got questions or want to connect? Join our Discord community
• If you are interested in custom solutions with edge deployment, please contact our sales team.

Citation

@article{liquidAI202624BA2B,
  author  = {Liquid AI},
  title   = {LFM2.5-24B-A2B: Built for Fast On-Device Agents},
  journal = {Liquid AI Blog},
  year    = {2026},
  note    = {www.liquid.ai/blog/lfm2-5-24b-a2b},
}

@article{liquidAI2026dspark,
  author = {Liquid AI},
  title = {LFM2.5-DSpark: Up to 3.2x Faster Inference from H100 to MacBook},
  journal = {Liquid AI Blog},
  year = {2026},
  note = {www.liquid.ai/blog/lfm2.5-dspark},
}