XHToken releases Spark-X2.6-4B GGUF for on-device use
Original title: XHToken/Spark-X2.6-4B-GGUF — new model trending #30 on Hugging Face
The 4B model keeps the Spark-X2.5 architecture and claims coding and agent gains from post-training alone.
XHToken published Spark-X2.6-4B, including a GGUF listing that was trending on Hugging Face. The card says it uses the same architecture as Spark-X2.5-4B and improves coding, agents, long context, multilingual ability and reasoning through post-training. It offers none, low and high thinking levels. The authors claim best-in-class results among similar-size open models and parity with some about-9B models; those figures are their own benchmarks. Quickstart notes name Transformers 4.57.1 and SGLang serving.
Unconfirmed: Benchmark leadership and parity with larger models are the authors’ claims, not independently verified.
- Thinking levels are none, low and high.
- The card compares inference with Qwen3.5-4B and Gemma-E4B.
- An example SGLang server sets context length to 1,048,576 tokens.
- The model id in the Transformers snippet is XHToken/Spark-X2.6-4B.
- Organizations
- XHToken
- AI models
- Spark-X2.6-4BSpark-X2.5-4B
From the original article
6 min read at huggingface.cotext-generation · 69,669 downloads · 162 likes · license apache-2. 0
By XHToken · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.