AI BriefingSparkLLM Team Releases Spark-X2.6-4B, a 4B On-Device Model Matching 9B Performance
·2026.10.21 09:00
Key point
The model supports a 1,048,576-token context and offers controllable thinking budgets to balance reasoning quality and latency.
Details
- SparkLLM Team released Spark-X2.6-4B, a compact 4B parameter on-device language model that serves as the successor to Spark-X2.5-4B.
- The model achieves state-of-the-art performance among similar-sized open-source models, matching or exceeding the capabilities of approximately 9B parameter models in agentic workflows, coding, and reasoning.
- It features a Controllable Thinking Budget with three settings (none, low, high) to dynamically adjust reasoning quality, latency, and inference cost.
- Deployment is supported across multiple frameworks including Transformers, SGLang, vLLM, MLX, llama.cpp, Ollama, and LM Studio, with specific image tags and version requirements provided for NVIDIA GPUs and Ascend NPUs.
- The model is distributed under the Apache 2.0 License and uses BF16 tensor types, with a base model identified as XHToken/Spark-X2.6-4B-Base.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.