XHToken/Spark-X2.5-4B

Spark-X2.5-4B is a compact, general-purpose language model intended for everyday tasks such as conversation, writing, translation, reasoning, coding, tool use, and agentic workflows. It uses a hybrid attention architecture combining full-attention and sliding-window layers, natively supports context windows of up to 1M tokens, and supports more than 200 languages. The model is compatible with hardware platforms including NVIDIA and Huawei and with inference frameworks such as vLLM, SGLang, and llama.cpp.

Appearances
4
Categories
1
First seen
17 September 2026
Latest
21 September 2026
Saved risers · this browser →← All current risersGet this page's badge ↗Open source ↗

Recorded measurements

Published observations only. Missing dates are not zero, and points do not establish what happened between observations. Each source and unit is shown separately.

Hugging Face model · downloads

Hugging Face models & datasets · 6 dated measurements

Rolling download counts, not lifetime totals or downloads gained between these dates.

Latest value shown: 30,772 downloads ·

Higher values appear higher on the chart. The scale covers the recorded values.

Hugging Face model: downloads recorded measurements6 dated points, 14 Sept 2026 to 21 Sept 2026. Values range from 21,336 to 30,772 downloads. Higher values appear higher. Open the exact values and published evidence below for the data table. Dots are not connected; missing dates are not zero.14 Sept 2026: 21,336 downloads16 Sept 2026: 25,650 downloads17 Sept 2026: 27,191 downloads18 Sept 2026: 28,347 downloads19 Sept 2026: 29,684 downloads21 Sept 2026: 30,772 downloads
View 6 exact dated values and published evidence
Exact values and their published evidence
Measured onValue (downloads)Published in
21,336
25,650
27,191
28,347
29,684
30,772

“Measured on” uses the original observation date. “Published in” links to the edition that reported it; the dates may differ.

01/

Appearance history

4 appearances
This page records every appearance of the same entity in the reviewed RiseFinder exports. Rankings keep the original category, period and measurement direction.

If you investigated this riser, tell us what helped—or what was missing.

Share feedback →
02/

Get email updates

Join the list

Get early signals on the websites, apps, repositories and startups gaining momentum. Join the list for Risefinder email briefings.

By subscribing, you ask us to store your email and send Risefinder updates. You can unsubscribe at any time. Privacy.

Want a sample first? Read past daily emails.