On Oct 19, XHToken released Spark-X2.6-4B, an on-device model with selectable thinking levels and stronger agent and coding results.
On Oct 19, XHToken released Spark-X2.6-4B, an on-device model with selectable thinking levels and stronger agent and coding results.
Hugging Face (XHToken) published this on Oct 19, 2026 (source: huggingface.co).
XHToken has released Spark-X2.6-4B on Hugging Face under Apache 2.0, an update to Spark-X2.5-4B with the same architecture and improved post-training. It says the model improves coding, agent, long-context, multilingual and reasoning tasks and matches or beats some 9B models, by its own benchmarks. The model offers three thinking levels (none, low and high), with GGUF, FP8, INT8 and INT4 versions.
Items older than 72 hours never go in Top stories and show their original date.