Models/LFM2.5 8B A1B DSpark

Text / LiquidAI

LFM2.5 8B A1B DSpark

LFM2.5-8B-A1B-DSpark is a speculative-decoding draft model from Liquid AI that adapts DSpark for the LFM2.5 architecture. It is designed to work with the target model LiquidAI/LFM2.5-8B-A1B, enabling faster decoding without degrading output quality. The draft model has 327.7M parameters in BF16 precision, a backbone of 5 full attention layers, a Markov head (rank 256), and a confidence head, with a block size of 9 and a vocabulary of 128,000. It is intended for use with SGLang and can run on Apple silicon via the Metal backend. The model card reports acceleration in SGLang decoding and on-device inference metrics, but these figures are provided as source claims and are not independently verified here.

EstimatedRepository opened Aug 10, 2026Source checked 9/24/2026Version: 5b285c82

LOCALRENTED GPUOPEN WEIGHTS

Freshness 70 out of 100. Repository age: 91 days old, Aug 10, 2026. Measures age, not quality.

Popularity 42 out of 100. 1,137 downloads in 30 days; 45 lifetime likes. Checked Nov 8, 2026. Measures adoption, not quality.

What do these scores mean?

Two ways to compare models. Neither score measures answer quality, benchmark performance, or whether a model fits your machine.

Freshness · release age

A new release starts at 100. The score halves every 180 days: 100 × 2^(−age in days / 180). When the release date is unavailable, a repository date is a fallback—not proof of when the model was released. Missing dates stay unscored.

Check the date source

Popularity · adoption

Downloads over the past 30 days carry 80% of the score; lifetime likes carry 20%. Both use a logarithmic scale, capped at 10 million downloads and 100,000 likes. If likes are unavailable, downloads carry all the weight. Missing download statistics stay unscored.

round(80 × min(1, log10(1 + downloads) / 7) + 20 × min(1, log10(1 + likes) / 5))

Check the statistics source

Scores as of Nov 9, 2026. Statistics checked Nov 8, 2026. Statistics are marked for refresh after 72 hours.Full scoring method

At a glance

Parameters: 0.33B
Architecture: qwen3
Context length: 128,000
License: lfm1.0
Disk space: 0.6 GB
Software: SGLang

View the model source

MY HARDWARE

Will it run on your machine?

Save your machine to see a personalized rating and its reasoning.

Add my hardware

Best for

Accelerating text generation with the LFM2.5-8B-A1B target model in SGLang
On-device inference on Apple silicon through the Metal backend
Speculative decoding for code generation tasks
Speculative decoding for mathematical reasoning tasks
Speculative decoding for general chat and instruction-following tasks

Tradeoffs

Designed as a draft model for LiquidAI/LFM2.5-8B-A1B and not a standalone generation model
Requires a build of SGLang with DSpark support for LFM2/LFM2-MoE targets
Licensed under lfm1.0 (other), which may impose restrictions on use
Performance depends on the target model and specific inference setup
Reported speedups and acceptance rates are from the model card and not independently tested here

Ways to run it

SGLang · Linux

Repository-specific command in publisher documentation; confirm dependencies and hardware in the source.

python -m sglang.launch_server  --model-path LiquidAI/LFM2.5-8B-A1B  --speculative-algorithm DSPARK  --speculative-draft-model-path LiquidAI/LFM2.5-8B-A1B-DSpark  --speculative-draft-attention-backend flashinfer  --disable-radix-cache --mem-fraction-static 0.75 --port 30000

Official instructions

Explore its uses

SAFETENSORSTEXT-GENERATION

Keep exploring

Text

Prism ML

Bonsai 2 · 27B

A compressed 27B-class reasoning model with publisher-provided GGUF packs and a dedicated llama.cpp fork for CUDA, Metal, and CPU.

Freshness 82 out of 100. Release age: 53 days old, Sep 17, 2026. Measures age, not quality.

MODEL SIZE27.36B

LOCALRENTED GPU

Released Sep 17, 2026

Text

DeepSeek

DeepSeek V4.1 Flash

An open-weight model for reasoning, coding, agents, and image understanding. DeepSeek V4.1 Flash combines a 552B backbone with sparse activation and up to one million tokens of context. Choose the original checkpoint or a community uncensored variant below.

Freshness 79 out of 100. Repository age: 60 days old, Sep 10, 2026. Measures age, not quality.

MODEL SIZE552B backbone · 8B active prefill / 16B active decode

LOCALRENTED GPU

Repository opened Sep 10, 2026

Text

Qwen

Qwen3.8

Explore 4 versions of Qwen3.8. Compare model sizes, packages, hardware requirements, and available setups.

4 versionsCompare sizes & formats

LOCALRENTED GPU

Compare versions →