Pricing
Input 2 CNY / 1M tokens
Output 8 CNY / 1M tokens
Text to Image -
Image to Image -
Text to Video -
Image to Video -
Capabilities
Input Types Text
Output Types Text
Function Calling Supported
Session Cache Supported
Fine-Tuning Supported
Prefix Cache Supported
Structured Output Not supported
Image Understanding Not supported
Text Generation Supported
Text to Image -
Image to Image -
Text to Video -
Image to Video -
Open Source Status Open source
Parameters
Last Updated 2026-07-25
Context Length 32K
Max Input Tokens 128K
Max Output Tokens 16K
Max Thought Tokens -
TPM 5000K
RPM 30K
Parameter Count -
Modalities Multimodal
Language Support Multilingual
Best For MoE
Resolution -
Video Duration -
Frame Rate -
Benchmarks
These benchmark scores summarize how the model performs across different tasks and evaluation sets.
AIME
39.2 Measures performance on difficult competition-level mathematics problems.
GPQA
80.7 Evaluates graduate-level science knowledge and reasoning ability.
LiveCodeBench
74.9 Benchmarks code understanding and real-world coding task performance.
Math500
90.2 Measures accuracy on a focused set of mathematics benchmark questions.
HumanEval
92 Evaluates code generation correctness on programming tasks.
MMLU Pro
85 Tests multidisciplinary knowledge, comprehension, and reasoning.
Coding Index
48.5 Summarizes overall coding capability across supported tasks.