Accédez aux meilleurs modèles AI via une API unique et unifiée. Tous les modèles sont compatibles OpenAI.
Kwaivgi Kling V3 — high-quality text-to-video and image-to-video. 3–15s, std/pro modes.
Kwaivgi Kling V3 Omni — multi-modal video generation with advanced subject control. 3–15s, std/pro modes.
Top-tier AI video generation. Text-to-video, image-to-video, video editing with audio. Cinema-quality output.
Fast AI video generation. Same quality as Seedance 2.0 with faster processing speed.
ByteDance flagship image generation model. Ultra-high quality text-to-image with photorealistic output.
Lightweight version of Seedream 5.0. High-quality text-to-image at lower cost.
Balanced image generation with broad style support and high fidelity.
OpenAI's latest image generation model. Supports multiple sizes, quality levels, and output formats.
OpenAI 2026 latest flagship, 1M context, all-around champion
GPT-5.4 Pro version, ultimate reasoning and code generation
GPT-5.4 lightweight version, fast and cost-effective
GPT-5.4 ultra lightweight, extremely low latency and cost
OpenAI code-specialized model, advanced programming and reasoning
Anthropic 2026 latest flagship, coding and reasoning fully upgraded
Anthropic 2026 most powerful flagship, autonomous agent task completion
Anthropic 2026 latest most powerful flagship, autonomous agent task completion
Claude lightweight version, fast response, most cost-effective
Claude main model, balancing capability and cost
Claude flagship model, complex task handling
Google 2026 most powerful flagship, code and reasoning breakthrough
Gemini 3.1 lightweight, optimized for speed and cost
Gemini 3.1 image generation, text-to-image capability
Google Gemini 3 Pro with image generation capabilities
Gemini 3 lightweight, fast and efficient
xAI latest reasoning model, fast deep thinking
xAI fast model, ultra-fast response
xAI code-focused model, strong programming capabilities
Qwen flagship model, strongest overall capabilities
Enhanced model, balancing capability and cost
Fast response model, suitable for high-concurrency scenarios
Zhipu latest GLM-5.1 flagship, advanced reasoning and generation
Zhipu fastest GLM-5 variant with 200K context
MiniMax latest flagship with 205K context, advanced reasoning
MiniMax advanced model with 205K context window
Xiaomi flagship AI model with 1M context window
Xiaomi multimodal AI model with 262K context
Qwen3.5 largest model, 397B parameters with MoE architecture
Qwen3 vision-language model, 235B parameters with 1M context
Qwen3 code-specialized model, strongest coding capabilities
Moonshot Kimi latest flagship with 262K context
Kimi deep reasoning model with chain-of-thought
Kimi K2 balanced model, strong general capabilities
DeepSeek 2026 top-tier flagship, 1M context, supports thinking & non-thinking modes, extended output up to 384K.
DeepSeek 2026 cost-effective flagship, 1M context, supports thinking & non-thinking modes. Successor to deepseek-chat/deepseek-reasoner.
DeepSeek V3.2 latest model, most cost-effective
DeepSeek V3.1 enhanced model, improved capabilities
DeepSeek V3.2 thinking model with extended chain-of-thought