SparkLLM is an open-source Spark-X 2.5 edge model, providing versions 4B and 1.7B.
SparkLLM is an open-source client-side model, Spark-X 2.5-4B and 1.7B, with a native context of up to 1 million tokens, covering more than 200 languages, and licensed under the Apache 2.0 license.
SparkLLM is an open-source, compact, general-purpose language model consisting of Spark-X2.5-4B and Spark-X2.5-1.7B.According to the official repository, the two models are designed for dialogue, writing, translation, reasoning, programming, tool calls, and agent workflows, with a native context of up to 1 million tokens and support for more than 200 languages.
Spark-X 2.5 employs a hybrid architecture combining full attention layers and sliding window attention layers. Download links for Hugging Face, ModelScope, and Ollam are provided, and it is compatible with inference frameworks such as vLLM, SGLang, llama.cpp, and MLX. The model series is licensed under the Apache 2.0 license.
refer to:Spark-X 2.5 Official Repository