Spark-X2.5#
Context Length: 1048576
Model Name: Spark-X2.5
Languages: en, zh
Abilities: chat, tools, reasoning, hybrid
Description: Spark-X2.5 is a compact general-purpose instruction model series with 1.7B and 4B checkpoints, native 1M-token context, reasoning, and tool-use support.
Specifications#
Model Spec 1 (pytorch, 1_7 Billion)#
Model Format: pytorch
Model Size (in billions): 1_7
Quantizations: none
Engines: vLLM, Transformers, SGLang
Model ID: XHToken/Spark-X2.5-1.7B
Model Hubs: Hugging Face, ModelScope
Execute the following command to launch the model, remember to replace ${quantization} with your
chosen quantization method from the options listed above:
xinference launch --model-engine ${engine} --model-name Spark-X2.5 --size-in-billions 1_7 --model-format pytorch --quantization ${quantization}
Model Spec 2 (pytorch, 4 Billion)#
Model Format: pytorch
Model Size (in billions): 4
Quantizations: none
Engines: vLLM, Transformers, SGLang
Model ID: XHToken/Spark-X2.5-4B
Model Hubs: Hugging Face, ModelScope
Execute the following command to launch the model, remember to replace ${quantization} with your
chosen quantization method from the options listed above:
xinference launch --model-engine ${engine} --model-name Spark-X2.5 --size-in-billions 4 --model-format pytorch --quantization ${quantization}
Model Spec 3 (pytorch, 1_7 Billion)#
Model Format: pytorch
Model Size (in billions): 1_7
Quantizations: Int8
Engines: Transformers
Model ID: XHToken/Spark-X2.5-1.7B-INT8
Model Hubs: Hugging Face, ModelScope
Execute the following command to launch the model, remember to replace ${quantization} with your
chosen quantization method from the options listed above:
xinference launch --model-engine ${engine} --model-name Spark-X2.5 --size-in-billions 1_7 --model-format pytorch --quantization ${quantization}
Model Spec 4 (pytorch, 4 Billion)#
Model Format: pytorch
Model Size (in billions): 4
Quantizations: Int8
Engines: Transformers
Model ID: XHToken/Spark-X2.5-4B-INT8
Model Hubs: Hugging Face, ModelScope
Execute the following command to launch the model, remember to replace ${quantization} with your
chosen quantization method from the options listed above:
xinference launch --model-engine ${engine} --model-name Spark-X2.5 --size-in-billions 4 --model-format pytorch --quantization ${quantization}
Model Spec 5 (fp8, 1_7 Billion)#
Model Format: fp8
Model Size (in billions): 1_7
Quantizations: FP8
Engines: vLLM, Transformers, SGLang
Model ID: XHToken/Spark-X2.5-1.7B-FP8
Model Hubs: Hugging Face, ModelScope
Execute the following command to launch the model, remember to replace ${quantization} with your
chosen quantization method from the options listed above:
xinference launch --model-engine ${engine} --model-name Spark-X2.5 --size-in-billions 1_7 --model-format fp8 --quantization ${quantization}
Model Spec 6 (fp8, 4 Billion)#
Model Format: fp8
Model Size (in billions): 4
Quantizations: FP8
Engines: vLLM, Transformers, SGLang
Model ID: XHToken/Spark-X2.5-4B-FP8
Model Hubs: Hugging Face, ModelScope
Execute the following command to launch the model, remember to replace ${quantization} with your
chosen quantization method from the options listed above:
xinference launch --model-engine ${engine} --model-name Spark-X2.5 --size-in-billions 4 --model-format fp8 --quantization ${quantization}
Model Spec 7 (ggufv2, 1_7 Billion)#
Model Format: ggufv2
Model Size (in billions): 1_7
Quantizations: Q4_K_M, Q8_0
Engines: vLLM, llama.cpp
Model ID: XHToken/Spark-X2.5-1.7B-GGUF
Model Hubs: Hugging Face, ModelScope
Execute the following command to launch the model, remember to replace ${quantization} with your
chosen quantization method from the options listed above:
xinference launch --model-engine ${engine} --model-name Spark-X2.5 --size-in-billions 1_7 --model-format ggufv2 --quantization ${quantization}
Model Spec 8 (ggufv2, 4 Billion)#
Model Format: ggufv2
Model Size (in billions): 4
Quantizations: Q4_K_M, Q8_0
Engines: vLLM, llama.cpp
Model ID: XHToken/Spark-X2.5-4B-GGUF
Model Hubs: Hugging Face, ModelScope
Execute the following command to launch the model, remember to replace ${quantization} with your
chosen quantization method from the options listed above:
xinference launch --model-engine ${engine} --model-name Spark-X2.5 --size-in-billions 4 --model-format ggufv2 --quantization ${quantization}