# XHToken releases Spark-X2.5, a 4B open LLM

> The Apache-2.0 instruction-tuned model targets developers who want a small, permissively licensed text generator.

Published by The Open Weights on Aug 24, 2026. Canonical: https://theopenweights.com/news/spark-x2-5-4b-abea

## Key facts

- Company: XHToken
- Model: Spark-X2.5-4B
- Version: X2.5
- Category: Text / LLM
- Modalities: Text / LLM
- License: Apache 2.0 (Open weights, commercial use allowed)
- Parameters: 4B
- Context window: 1.0M tokens
- Architecture: Spark2_5ForCausalLM
- Precision: BF16
- Weights size: 8.2 GB
- Significance: minor
- Published: 2026-08-24
- Last verified: 2026-09-23
- Hugging Face: https://huggingface.co/XHToken/Spark-X2.5-4B
- Canonical URL: https://theopenweights.com/news/spark-x2-5-4b-abea

XHToken has published Spark-X2.5-4B, a four-billion-parameter instruction-tuned language model released under the permissive Apache-2.0 license. The model is available now on [Hugging Face](https://huggingface.co/XHToken/Spark-X2.5-4B), where it is listed as a text-only, dense (non-MoE) system.

At 4B parameters, Spark-X2.5 sits in the increasingly crowded small-model tier, where compact instruction-tuned LLMs are prized for running on modest hardware, fine-tuning cheaply, and serving latency-sensitive applications. The Apache-2.0 terms are the notable detail here: they allow commercial use, modification, and redistribution with minimal friction, which matters to teams wary of the usage restrictions attached to some other open-weight releases.

## What we know

- Roughly 4B parameters, dense architecture
- Instruction-tuned for chat and task following
- Apache-2.0 license, text-only modality
- Distributed via Hugging Face

The release record leaves several specifics unstated, including context length and benchmark results, so buyers will want to test the model against their own workloads before committing. As an initial entry in the Spark-X line, it establishes a baseline that future revisions can build on.

Why it matters: the value of a small model like this rests less on headline scores than on the combination of size and license. For developers who need a self-hostable, commercially usable 4B model, Spark-X2.5 is worth a look — and worth benchmarking carefully given how much the record leaves open.

## Get the model

- [Hugging Face](https://huggingface.co/XHToken/Spark-X2.5-4B)

## Sources

- [XHToken/Spark-X2.5-4B](https://huggingface.co/XHToken/Spark-X2.5-4B) — Hugging Face, Aug 24, 2026

---
Source: The Open Weights (https://theopenweights.com/). Aggregated and written by Claude, curated by humans. Cite as: "XHToken releases Spark-X2.5, a 4B open LLM", The Open Weights, Aug 24, 2026, https://theopenweights.com/news/spark-x2-5-4b-abea