# What Liquid DSpark’s “open” release actually permits

Agent contract: 1.2.0

The weights are downloadable and the acceleration technique is inspectable. The license still places a commercial-use condition on larger organizations.

Canonical: https://brightaifuture.com/guides/liquid-dspark-open-weights-license
Format: guide
Source publication: Not established
Bright publication: 2026-09-25
Substantive update: None recorded
Evidence and review: Primary-source review completed 2026-09-25; Bright did not reproduce the cited research.

## Answer first

Liquid released a 279.5-million-parameter experimental draft model designed to work with LFM2.5-VL-3B during speculative decoding. Its weights and integrations are available, but its LFM Open License v1.0 requires a separate commercial license for an organization with at least $10 million in annual revenue. “Downloadable weights” therefore does not mean unrestricted commercial use or fully open source.

## The drafter and the base model are a pair

DSpark does not replace the 3-billion-parameter vision-language model. It proposes candidate output tokens; the base model verifies them and preserves the final model distribution. The 279.5-million-parameter drafter is labeled experimental and must be paired with LFM2.5-VL-3B.

Liquid reports up to 3.13× faster decoding on an Apple M5 Max and 2.66× on an NVIDIA H100. Those are company benchmarks for the token-generation stage. They do not establish the same improvement for image encoding, prompt processing, every sampling mode, or an entire user request.

Evidence [liquid-dspark-post]: Introducing LFM2.5-VL-3B-DSpark. Liquid AI’s release explanation and company-reported decoding benchmarks. https://www.liquid.ai/blog/lfm2-5-vl-dspark

Evidence [liquid-dspark-card]: LFM2.5-VL-3B-DSpark model card. Release status, paired base model, parameter count, integrations and experimental limitations. https://huggingface.co/LiquidAI/LFM2.5-VL-3B-DSpark

## Code, weights and data answer different questions

The public model repository provides downloadable weights and documentation, while the release names llama.cpp, MLX-VLM and SGLang integrations. Each runtime has its own code and terms. None of that, by itself, publishes the training data or a complete reproducible training recipe.

The useful inspection is artifact by artifact: the drafter weights, base-model weights, runtime code, license, model card, evaluation method and training-data account. One open component does not transfer its permissions to the others.

Evidence [liquid-dspark-card]: LFM2.5-VL-3B-DSpark model card. Release status, paired base model, parameter count, integrations and experimental limitations. https://huggingface.co/LiquidAI/LFM2.5-VL-3B-DSpark

Evidence [liquid-dspark-license]: LFM2.5-VL-3B-DSpark LICENSE. The DSpark repository’s copy of LFM Open License v1.0, including its commercial-use threshold. https://huggingface.co/LiquidAI/LFM2.5-VL-3B-DSpark/blob/main/LICENSE

## Read the license before the launch language

The repository’s own LICENSE is the operative document Bright found for the DSpark weights. It allows many uses but defines a commercial-use restriction: organizations with annual revenue of $10 million or more need a separate license from Liquid AI. That condition matters even when release copy uses broader language about access or use.

Bright therefore describes DSpark as an open-weight release under a source-available license with a commercial threshold. It does not call the model open source, certify the license, or infer rights for data and third-party runtime code.

Evidence [liquid-dspark-license]: LFM2.5-VL-3B-DSpark LICENSE. The DSpark repository’s copy of LFM Open License v1.0, including its commercial-use threshold. https://huggingface.co/LiquidAI/LFM2.5-VL-3B-DSpark/blob/main/LICENSE

Evidence [liquid-dspark-post]: Introducing LFM2.5-VL-3B-DSpark. Liquid AI’s release explanation and company-reported decoding benchmarks. https://www.liquid.ai/blog/lfm2-5-vl-dspark

## Limitations

The speed figures are Liquid AI’s decode-only benchmarks and were not independently reproduced by Bright.

MLX support in the cited release is limited to greedy sampling, and performance depends on hardware, runtime, prompt and generation settings.

Bright is reporting the repository terms checked on September 25, 2026, not providing legal advice or certifying license compatibility.

Public weights and inference integrations do not establish that training data, training code or a complete training recipe are available.

## Provenance and history

{
  "provenance": {
    "publisher": "Bright",
    "role": "Source-backed evergreen explanation; not a new scientific result"
  }
}

## Original sources

- [LFM2.5-VL-3B-DSpark LICENSE](https://huggingface.co/LiquidAI/LFM2.5-VL-3B-DSpark/blob/main/LICENSE)
- [LFM2.5-VL-3B-DSpark model card](https://huggingface.co/LiquidAI/LFM2.5-VL-3B-DSpark)
- [Introducing LFM2.5-VL-3B-DSpark](https://www.liquid.ai/blog/lfm2-5-vl-dspark)

## Continue exploring

- [Explore Open Models](https://brightaifuture.com/open-models)