{"data":[{"capacity_tpm":10000000,"context_length":131072,"created":1774337486,"datacenters":[{"country_code":"CN"}],"deprecation_date":" ","description":"DeepSeek-V3.2 has officially reached stable release. Engineered to strike the perfect balance between reasoning depth and token efficiency, this model is ideal for everyday applications, ranging from standard QA to general-purpose agent workflows.","discount_to_user":0,"hugging_face_id":"deepseek-ai/DeepSeek-V3.2","id":"deepseek-v3.2-intl","input_modalities":["text"],"max_output_length":65536,"name":"Baidu Qianfan: DeepSeek-V3.2","openrouter":{"slug":"baidu-qianfan/deepseek-v3.2"},"output_modalities":["text"],"pricing":{"completion":"0.00000042","image":"0","input_cache_read":"0.000000028","prompt":"0.00000028","request":"0"},"quantization":"fp8","supported_features":["tools","structured_outputs","reasoning"],"supported_sampling_parameters":["temperature","top_p","stop","seed","max_tokens"]},{"context_length":202752,"created":1774337486,"datacenters":[{"country_code":"CN"}],"deprecation_date":"","description":"GLM-5 is Zhipu’s next-generation flagship base model, purpose-built for agentic engineering. It delivers robust performance for complex systems engineering and long-horizon agent workflows.","hugging_face_id":"zai-org/GLM-5","id":"glm-5","input_modalities":["text"],"max_output_length":131072,"name":"Baidu Qianfan: GLM-5","openrouter":{"slug":"baidu-qianfan/glm-5"},"output_modalities":["text"],"pricing":{"completion":"0.00000224","image":"0","input_cache_read":"0.00000014","prompt":"0.0000007","request":"0"},"quantization":"fp8","supported_features":["tools","structured_outputs","reasoning"],"supported_sampling_parameters":["temperature","top_p","frequency_penalty","presence_penalty","repetition_penalty","stop","seed","max_tokens"]},{"context_length":65536,"created":1775057354,"datacenters":[{"country_code":"CN"}],"deprecation_date":"2026-05-28","description":"Qianfan-OCR-Fast is a domain-specific multimodal large model purpose-built for OCR. By leveraging specialized OCR training data while preserving versatile multimodal intelligence, it provides a powerful performance upgrade over Qianfan-OCR.","hugging_face_id":"","id":"qianfan-ocr-fast","input_modalities":["text","image"],"max_output_length":28672,"name":"Baidu Qianfan: Qianfan-OCR-Fast","openrouter":{"slug":"baidu-qianfan/qianfan-ocr-fast"},"output_modalities":["text"],"pricing":{"completion":"0.00000281","image":"0","input_cache_read":"0","prompt":"0.00000068","request":"0"},"quantization":"fp8","supported_features":[],"supported_sampling_parameters":["temperature","top_p","frequency_penalty","presence_penalty","repetition_penalty","stop","seed","max_tokens"]},{"capacity_tpm":100000000,"context_length":1048576,"created":1777385869,"datacenters":[{"country_code":"CN"}],"deprecation_date":" ","description":"DeepSeek-V4 features an ultra-long context window of one million tokens. It stands as a leader in both the domestic and open-source sectors, delivering top-tier performance in agentic capabilities, world knowledge, and reasoning.DeepSeek-V4-Pro features a total of 1.6T parameters with 49B activated parameters.","discount_to_user":0.75,"hugging_face_id":"deepseek-ai/DeepSeek-V4-Pro","id":"deepseek-v4-pro","input_modalities":["text"],"max_output_length":393216,"name":"Baidu Qianfan: DeepSeek-V4-Pro","openrouter":{"slug":"baidu-qianfan/deepseek-v4-pro"},"output_modalities":["text"],"pricing":{"completion":"0.00000338","image":"0","input_cache_read":"0.00000014","prompt":"0.00000169","request":"0"},"quantization":"fp8","supported_features":["tools","structured_outputs","reasoning"],"supported_sampling_parameters":["temperature","top_p","stop","max_tokens"]},{"capacity_tpm":30000000,"context_length":1048576,"created":1777385869,"datacenters":[{"country_code":"CN"}],"deprecation_date":" ","description":"DeepSeek-V4-Flash is the lightweight, high-efficiency edition of the V4 series. While maintaining the 1M ultra-long context capability, it utilizes a smaller parameter count and activation scale to provide faster and more cost-effective API services.","discount_to_user":0,"hugging_face_id":"deepseek-ai/DeepSeek-V4-Flash","id":"deepseek-v4-flash","input_modalities":["text"],"max_output_length":131072,"name":"Baidu Qianfan: DeepSeek-V4-Flash","openrouter":{"slug":"baidu-qianfan/deepseek-v4-flash"},"output_modalities":["text"],"pricing":{"completion":"0.00000028","image":"0","input_cache_read":"0.000000028","prompt":"0.00000014","request":"0"},"quantization":"fp8","supported_features":["tools","structured_outputs","reasoning"],"supported_sampling_parameters":["temperature","top_p","stop","max_tokens"]},{"capacity_tpm":10000000,"context_length":202752,"created":1778759134,"datacenters":[{"country_code":"CN"}],"deprecation_date":"","description":"GLM-5.1 is Z.ai’s next-generation flagship foundation model, featuring a significant leap in long-horizon task performance. It is capable of autonomous operation for up to 8 hours, delivering closed-loop, engineering-grade results. Its overall performance is on par with Claude 4.6 Opus.","discount_to_user":0,"hugging_face_id":"zai-org/GLM-5.1","id":"glm-5.1","input_modalities":["text"],"max_output_length":131072,"name":"Baidu Qianfan: GLM-5.1","openrouter":{"slug":"baidu-qianfan/glm-5.1"},"output_modalities":["text"],"pricing":{"completion":"0.0000044","image":"0","input_cache_read":"0.00000026","prompt":"0.0000014","request":"0"},"quantization":"fp8","supported_features":["tools","structured_outputs","reasoning"],"supported_sampling_parameters":["temperature","top_p","frequency_penalty","presence_penalty","repetition_penalty","stop","max_tokens"]},{"capacity_tpm":20000000,"context_length":262144,"created":1778759134,"datacenters":[{"country_code":"CN"}],"deprecation_date":"","description":"Kimi K2.6 is the latest and most intelligent model in the Kimi portfolio. It delivers comprehensive enhancements across general agentic capabilities, coding, and visual understanding. It achieves industry-leading results on elite benchmarks—including Humanity’s Last Exam (a PhD-level assessment), SWE-Bench Pro (the gold standard for real-world software engineering capabilities), and DeepSearchQA (which evaluates an agent's deep retrieval proficiency).","discount_to_user":0.39,"hugging_face_id":"moonshotai/Kimi-K2.6","id":"kimi-k2.6","input_modalities":["text","image"],"max_output_length":262144,"name":"Baidu Qianfan: Kimi-K2.6","openrouter":{"slug":"baidu-qianfan/kimi-k2.6"},"output_modalities":["text"],"pricing":{"completion":"0.000004","image":"0","input_cache_read":"0.00000016","prompt":"0.00000095","request":"0"},"quantization":"fp4","supported_features":["tools","structured_outputs","reasoning"],"supported_sampling_parameters":["temperature","top_p","frequency_penalty","presence_penalty","stop","max_tokens"]},{"capacity_tpm":80000000,"context_length":1048576,"created":1781783656,"datacenters":[{"country_code":"CN"}],"deprecation_date":"","description":"GLM-5.2 is a flagship model engineered for the era of long-horizon tasks. Supporting a 1M context window, it delivers unparalleled stability in long-horizon task execution and more reliable adherence to engineering standards, further boosting success rates across complex development scenarios. In a single, seamless workflow, GLM-5.2 orchestrates the entire development lifecycle—from requirements gathering to multi-platform, deployment-ready outputs.","discount_to_user":0.65,"hugging_face_id":"zai-org/GLM-5.2","id":"glm-5.2","input_modalities":["text"],"max_output_length":131072,"name":"Baidu Qianfan: GLM-5.2","openrouter":{"slug":"baidu-qianfan/glm-5.2"},"output_modalities":["text"],"pricing":{"completion":"0.0000044","image":"0","input_cache_read":"0.00000026","prompt":"0.0000014","request":"0"},"quantization":"fp8","supported_features":["tools","structured_outputs","reasoning"],"supported_sampling_parameters":["temperature","top_p","frequency_penalty","presence_penalty","repetition_penalty","stop","max_tokens"]},{"context_length":1048576,"created":1781783656,"datacenters":[{"country_code":"CN"}],"deprecation_date":"","description":"MiMo-V2.5 is a native omni-modal foundation model engineered with powerful agentic capabilities, delivering unified understanding across text, images, video, and audio within a single, cohesive architecture. Built upon the high-efficiency MiMo-V2-Flash backbone and augmented with dedicated vision and audio encoders, MiMo-V2.5 excels in multimodal perception, long-context reasoning, and complex agentic workflows. ","discount_to_user":0.2,"hugging_face_id":"XiaomiMiMo/MiMo-V2.5","id":"mimo-v2.5","input_modalities":["text"],"max_output_length":131072,"name":"Baidu Qianfan: MiMo-V2.5","openrouter":{"slug":"baidu-qianfan/mimo-v2.5"},"output_modalities":["text"],"pricing":{"completion":"0.00000028","image":"0","input_cache_read":"0.0000000028","prompt":"0.00000014","request":"0"},"quantization":"fp8","supported_features":["tools","structured_outputs","reasoning"],"supported_sampling_parameters":["temperature","top_p","frequency_penalty","presence_penalty","repetition_penalty","stop","max_tokens"]},{"capacity_tpm":1,"context_length":1048576,"created":1777385869,"datacenters":[{"country_code":"CN"}],"deprecation_date":" ","description":"DeepSeek-V4-Flash-0731 is the official release of DeepSeek-V4-Flash, superseding the preview version, with substantially enhanced agentic capabilities. It has the same model structure as DeepSeek-V4-Flash-DSpark, i.e. it comes with a speculative decoding module attached.","discount_to_user":0,"hugging_face_id":"deepseek-ai/DeepSeek-V4-Flash-0731","id":"deepseek-v4-flash-0731","input_modalities":["text"],"max_output_length":131072,"name":"Baidu Qianfan: DeepSeek-V4-Flash-0731","openrouter":{"slug":"baidu-qianfan/deepseek-v4-flash-0731"},"output_modalities":["text"],"pricing":{"completion":"0.00000028","image":"0","input_cache_read":"0.000000028","prompt":"0.00000014","request":"0"},"quantization":"fp8","supported_features":["tools","structured_outputs","reasoning"],"supported_sampling_parameters":["temperature","top_p","stop","max_tokens"]},{"capacity_tpm":1000000,"context_length":262144,"created":1783091351,"datacenters":[{"country_code":"CN"}],"deprecation_date":"","description":"Hy3 is a 295B-parameter Mixture-of-Experts (MoE) model with 21B active parameters and 3.8B MTP layer parameters, developed by the Tencent Hy Team.","discount_to_user":0,"hugging_face_id":"tencent/Hy3","id":"hy3","input_modalities":["text"],"max_output_length":262144,"name":"Baidu Qianfan: Hy3","openrouter":{"slug":"baidu-qianfan/hy3"},"output_modalities":["text"],"pricing":{"completion":"0.000000528","image":"0","input_cache_read":"0.000000033","prompt":"0.000000132","request":"0"},"quantization":"fp8","supported_features":["tools","structured_outputs","reasoning"],"supported_sampling_parameters":["temperature","top_p","frequency_penalty","presence_penalty","repetition_penalty","stop","max_tokens"]}]}