Compare capabilities, context and current prices. Find the right fit for every task.
CNY ¥ · One entry per family. Price, capabilities and official badges apply to the shown variant. Sort within company and billing unit.
GPT Image 2 creates images from text prompts for product visuals, concepts, illustrations and poster assets. Compare the family’s routes and specifications, then use the selected route’s parameters, price and unique model ID.
GPT Image 2.5 Flare focuses on fast everyday image creation, including concept drafts, social visuals and iterative exploration. Its routes and prices are separate from Sunburst; check route details for available dimensions and quality options.
GPT-5.6 Luna is designed for cost-sensitive, high-volume tasks such as classification, extraction and concise answers. This route requires an output budget of at least 16 tokens and has separate long-context price bands.
GPT-6 Astra targets demanding reasoning, agentic coding and professional knowledge work, including multi-step planning and long-document analysis. Standard and long-context inputs have separate price bands.
GPT-5.6 Terra balances capability and cost for everyday coding, content processing and business automation. It is a practical option for sustained workloads with a defined budget.
The H3 2K route is available for short-video creation from text prompts. The current playground configuration allows up to 15 seconds and 9 reference images. Check route parameters for reference-video/audio support and dimensions.
Seedance 2.0 supports text, image and multimodal reference video generation at 720p, 480p, 1080p for 4–15 seconds. Text and reference-image generation is billed per second at the selected resolution. Reference-video requests add only the positive difference between their metered task price and the generation fee. Generation is not charged twice. See route parameters for supported reference formats and limits.
Seedance 2.5 supports text, image and multimodal reference video generation at 720p, 480p, 1080p for 4–30 seconds. Text and reference-image generation is billed per second at the selected resolution. Reference-video requests add only the positive difference between their metered task price and the generation fee. Generation is not charged twice. See route parameters for supported reference formats and limits.
Seedance 2.0 Fast supports text, image and multimodal reference video generation at 720p, 480p for 4–15 seconds. Text and reference-image generation is billed per second at the selected resolution. Reference-video requests add only the positive difference between their metered task price and the generation fee. Generation is not charged twice. See route parameters for supported reference formats and limits.
Seedance 2.0 Mini supports text, image and multimodal reference video generation at 720p, 480p for 4–15 seconds. Text and reference-image generation is billed per second at the selected resolution. Reference-video requests add only the positive difference between their metered task price and the generation fee. Generation is not charged twice. See route parameters for supported reference formats and limits.
Gemini 3 Pro Image Preview is a preview image-creation model for visual concepts, text-rich designs and more complex image prompts. Supported sizes, quality settings and input formats depend on this preview route.
Gemini 3.1 Flash Image Preview suits iterative image workflows, from quick drafts and illustrations to repeated visual refinements. This preview version has independent route pricing and a unique model ID.
Gemini Nano Banana 2.1 targets efficient image generation and conversational visual creation, including character assets, product visuals and text-rich designs. Check the selected route for resolution, reference-image support and billing.
Gemini 3.5 Flash is an option for everyday Q&A, text processing and development assistance in budget-conscious workflows. It is served as a distinct version with route-specific parameters and prices.
Gemini 3.1 Pro Preview targets more complex reasoning, coding and long-document work. This preview version may evolve upstream, and standard versus long-context inputs have separate price bands.
DeepSeek V4.1 Flash suits frequent everyday tasks such as Q&A, coding assistance and extraction. The undated route shares a family with dated variants; pricing and the hosted version remain route-specific.
DeepSeek V4 Pro targets deeper reasoning and coding work, including complex problem decomposition and professional document analysis. Compare it with Flash options according to workload and budget.
GLM-5.3 Flash focuses on efficient, budget-conscious workloads such as frequent Q&A, extraction and everyday development assistance. It is priced separately from GLM-5.3; multimodal input availability is route-specific.
GLM-5.3 is a Zhipu text model for Chinese knowledge work, coding, long-document analysis and multi-step reasoning. Input, output and cache reads are priced separately.
Kimi K3 targets long-horizon coding, knowledge work and deep reasoning. It suits development analysis, synthesis across multiple documents and planning complex workflows.
Kimi K2.6 suits long-text Q&A, development assistance and agent workflows. It is listed separately from K3 so you can compare route parameters, token prices and workload requirements.
Claude Opus 5.5 targets demanding coding, reasoning and professional document work, including analysis and implementation across longer workflows. Cache reads and writes follow the route’s published prices.
Claude Fable 5 is an option for demanding analysis, software development and multi-step professional work. It has distinct input, output and cache prices for workload and budget comparisons.
Claude Sonnet 5 suits everyday development and professional text work, including code changes, document organization and business automation. Check the route for available tools and context limits.
A Claude Code-specific route for Claude Sonnet 4.6, intended for code understanding and development assistance in that client. This route is restricted to Claude Code rather than general third-party API use.
A Claude Code-specific route for Claude Opus 4.6, suited to complex code analysis and multi-step development in that client. Claude Code restrictions and the route’s own prices apply.
号池大语言模型线路，非官方线路。支持参数及售价以本线路配置为准；同家族不同分组分别计价。
号池大语言模型线路，非官方线路。支持参数及售价以本线路配置为准；同家族不同分组分别计价。
号池大语言模型线路，非官方线路。支持参数及售价以本线路配置为准；同家族不同分组分别计价。
号池大语言模型线路，非官方线路。支持参数及售价以本线路配置为准；同家族不同分组分别计价。
号池大语言模型线路，非官方线路。支持参数及售价以本线路配置为准；同家族不同分组分别计价。
