Place des modèles

deepseek-v4.1-flash

DeepSeek

DeepSeek-V4.1-Flash 是 DeepSeek 当前主力模型，官方 API 模型名为 deepseek-flash。支持 1M 上下文与最大 384K 输出，默认可切换思考/非思考模式，具备图像理解、工具调用、JSON Output，并兼容 OpenAI 与 Anthropic API。

qwen3.8-flash

Qwen

The Flash tier of the Qwen3.8 series: multimodal reasoning for coding and agent workflows, with understanding of documents, codebases and long videos.

glm-5.3-flash

Z.AI

The Flash tier of the GLM-5.3 series: natively multimodal, built for efficient coding and long-chain agents, with stable long-context performance for production applications.

qwen3.8-max

Qwen

The flagship reasoning model of the Qwen3.8 series, accepting multimodal input for complex reasoning, visual understanding, coding and agent workflows.

kimi-k3

Moonshot

A multimodal reasoning model for complex coding, knowledge work, and long-running agent tasks. It works across large codebases, calls tools, and debugs, using images, logs, and test output to refine its output.

deepseek-v4-pro-0813

DeepSeek

The DeepSeek V4 flagship, with the series' strongest deep reasoning and code generation, suited to whole-codebase analysis, multi-step automation and demanding instructions.

glm-5.3

Z.AI

Flagship-tier model for complex software engineering and long-horizon agent tasks, with stable execution across multi-turn tool calls and large-scale codebase refactoring.

glm-5.2

Z.AI

GLM 5.2 is a large-scale reasoning model with a 1M-token context window, suited for long-horizon agents, repo-level coding, and multi-step automation.

kimi-k2.7-code

Moonshot AI

Coding-focused variant in the Kimi K2 family. Accepts text and image inputs, with reasoning on by default. Suited for long-horizon coding and agentic workflows.

seedream-5.0-pro

ByteDance

Seedream-5.0-pro是字节跳动发布的最新图像创作模型。该模型将图像创作推进到可控生产的新阶段，本次更新的主要亮点是编辑更可控、生产更落地、效果更自然。

seedance-2.0

ByteDance

Standard tier of the Seedance family. Supports text-to-video, image-to-video, and reference-to-video, with strong character, style, and camera consistency. Built for ads and short-form creative.

doubao-seed-2.0-mini

ByteDance

Lightweight tier in the Seed 2.0 family, tuned for high-throughput, low-latency tasks with adjustable reasoning levels. Fits batch processing, moderation, and classification.

seedance-2.0-mini

ByteDance

Seedance 2.0 mini is a new-generation, cost-effective video generation model for a broader range of needs. It keeps competitive quality while bringing video generation to lower-barrier, higher-frequency and larger-scale use cases.

seedance-2.0-fast

ByteDance

Lightweight Seedance 2.0 variant tuned for speed and cost rather than peak fidelity, with text-to-video, first/last frame control, and multi-reference inputs.

doubao-seed-2.0-mini-260428

ByteDance

Lightweight tier in the Seed 2.0 family, tuned for high-throughput, low-latency tasks with adjustable reasoning levels. Fits batch processing, moderation, and classification.

doubao-seed-2.0-pro

ByteDance

Flagship of the Seed 2.0 series for complex reasoning and long-horizon agent workflows with reliable tool use.

doubao-seed-2.0-pro-260215

ByteDance

Flagship of the Seed 2.0 series for complex reasoning and long-horizon agent workflows with reliable tool use.

qwen3.7-max

Qwen

Flagship agent-centric reasoning model with a 1M context window, excelling at coding, productivity, and long-horizon autonomous tasks.

doubao-seed-2.1-pro

ByteDance

From ByteDance, supports 文本 et 多模态 and more—call it via the standard API.

doubao-seed-evolving

ByteDance

From ByteDance, supports 文本 et 多模态 and more—call it via the standard API.

doubao-seed-2.0-lite

ByteDance

Lightweight tier of the Seed 2.0 family tuned for low-latency agent, coding, and GUI workloads.

doubao-seed-2.0-lite-260428

ByteDance

Lightweight tier of the Seed 2.0 family tuned for low-latency agent, coding, and GUI workloads.

kimi-k2.6

Moonshot AI

Latest multimodal model in the K2 series, built for long-horizon coding, code-driven UI/UX generation, and multi-agent orchestration.

minimax-m3

MiniMax

MiniMax\\'s first multimodal release, accepting image and video input. Long-context inference runs cheaper and faster, and it handles multi-step agent workflows and computer-use scenarios.

glm-5.1

Z.AI

Long-task autonomous coding model

deepseek-v4-flash-0731

DeepSeek

A faster, lower-cost version of the DeepSeek V4 series that keeps the main model's reasoning and coding foundations, suited to high-concurrency chat, coding assistance and latency-sensitive agents.

seedance-2.5

ByteDance

Seedance 2.5 是豆包大模型团队推出的新一代多模态视频创作模型，已迈入「长叙事 × 多素材 × 强编辑 × 多语言」的工业级新阶段。

qwen3.7-plus

Qwen

Qwen 3.7 Plus is a mid-tier multimodal model that reads screens, operates GUIs, and navigates mobile apps end-to-end. Suited for agent workflows and tool calling.

hy3-preview

Tencent

Hy3 preview in high-reasoning mode, built for complex problem solving and multi-step agent workflows that need dependable execution.

mimo-v2.5-pro

Xiaomi

Flagship coding and agent model that sustains thousand-tool-call autonomous workflows for complex software engineering.

hailuo-2.3

MiniMax

It supports Vincentian and Tusheng videos, can output 6-10 seconds, 768p/1080p images, supports fine camera movement command control, has smooth movement and stable subject, and is suitable for short video creative content.

wan2.7-t2v

Alibaba

Wan2.7-T2V delivers fully upgraded performances: nuanced, natural emotion in dramatic scenes, hard-hitting action sequences, and more dramatic, rhythmic shot transitions.

hailuo-02

MiniMax

It supports Vincentian and Tusheng videos, can output 6-10 seconds, 512p/768p/1080p multi-level image quality, supports camera movement command control, and flexibly balances image quality and cost.

wan2.6-t2v

Alibaba

Generate 720p/1080p video from text prompts and come with synchronized audio tracks. The picture has movie-level aesthetics and complex motion performance, suitable for creative video creation.

qwen-image-edit

Qwen

The first image editing model in the Qwen series, extending Qwen-Image's text rendering to editing tasks. Supports precise bilingual (Chinese and English) text editing, both appearance and semantic edits, and strong results across benchmarks.

qwen-image-2.0

Qwen

Unifies generation and editing in one model; native 2K; professional typography and infographics.

qwen-image-2.0-pro

Qwen

Fuses generation and editing; more professional text rendering and richer realistic textures and scenes.

seedream-4.0

ByteDance

New unified architecture combining generation and editing; knowledge-based output and reference consistency; up to 4K.

seedream-4.5

ByteDance

Broad upgrade via model scaling; sharper subject detection, stricter reference fidelity, stronger dense text rendering.

seedream-5.0-lite

ByteDance

Adds deep reasoning; better interprets complex prompts with common-sense knowledge; stronger edit consistency.

minimax-m2.7-highspeed

MiniMax

Speed-optimized variant of M2.7 with identical outputs, tuned for low-latency coding and agent tool-calling workloads.

qwen3.7-flash

Qwen

The lightweight vision-language reasoning model of the Qwen3.7 series, strong at object recognition, spatial understanding and real-world visual perception, built for multimodal agents, visual coding, search and computer use.

minimax-m2.5-highspeed

MiniMax

A high-throughput variant of M2.5 with matching quality at roughly triple the speed, tuned for coding and agentic tool use under latency-sensitive workloads.

mimo-v2.5

Xiaomi

Mid-tier V2.5 model with native image, audio, and video understanding, built for agent workflows and coding at lower cost than Pro.

seed-1.8

ByteDance

Agent-oriented foundation model tuned for tool calling and complex instruction following, with use cases spanning GUI agents, search agents, and multi-step task orchestration.

doubao-seed-2.0-code-preview-260215

ByteDance

Code-focused preview model for agent workflows, with competitive SWE-Bench and LiveCodeBench scores, aimed at autonomous coding agents and multi-language software tasks.

qwen3.5-flash

Alibaba

Lightweight mid-tier variant tuned for low-latency agent workflows with coding and tool-calling support.

qwen3.6-35b-a3b

Alibaba

Open-weight coding model tuned for agentic terminal tasks and repo-scale reasoning with low inference cost.

qwen3.5-27b

Alibaba

Mid-tier Qwen model tuned for multi-step reasoning and agent workflows, compact enough for private and on-prem deployment.

qwen3.6-27b

Alibaba

Thinking-mode variant of the dense open-weight model that outputs step-by-step reasoning; at 27B it surpasses the prior open-weight 397B-A17B flagship across coding, math and multi-step reasoning benchmarks.

qwen3.5-plus

Qwen

Multimodal model from the Qwen3.5 line with toggleable thinking mode, tuned for image and video understanding, document parsing, and multimodal agents.

qwen3.5-122b-a10b

Alibaba

Open-weight mid-tier MoE text model with long chain-of-thought reasoning, strong on function-calling and multi-step planning benchmarks for self-hosted agentic workflows.

qwen3-vl-plus

Qwen

Plus-tier vision-language model in the Qwen3-VL line, handling text, image and video inputs with strengths in document parsing, video understanding, spatial grounding and agent tool use.

glm-4.6

Z.AI

Text-only model in the GLM-4 family with improved coding benchmarks and tool calls during reasoning, suited for software development and agent workflows.

Kling-O1

Kling

Kling O1 是为创作者打造的全新创意引擎，解锁无限创意可能。基于多模态视觉语言（MVL）概念，使用自然语言结合视频、图片、元素等多模态描述，精准理解您的创作意图。

Kling-3.0-Omni

Kling

可灵3.0 Omni是旗舰模型,提供卓越的视频质量、先进的语义理解和增强的伪影修正。基于视频3.0的基础,Omni为专业创作者提供极致品质,具有更精细的细节、更流畅的运动动态和更高的成功率。

Kling-3.0

Kling

可灵视频 3.0 配备革命性智能分镜系统,自动编排多镜头序列,实现电影级转场效果。单次生成最长15秒视频,具备增强的角色一致性、原生级文字叠加和多语言音频混合能力。

Kling-2.6

Kling

VIDEO 2.6 模型一次生成视觉画面、自然配音、匹配音效和环境氛围，连接「声音」与「视觉」的世界。无论是输入文字还是上传图片，都能即时创建一个完整的动态视频，包含声音、节奏和沉浸感。

Kling-2.5

Kling

Kling 2.5 Turbo 针对速度和性价比进行了优化，非常适合大批量视频生成需求。在显著缩短生成时间的同时保持良好的质量。

Kling-2.1

Kling

Kling 2.1 显著提升了语义响应、动态质量和画面质感。是稳定、高质量视频生成的首选。

Kling-2.0

Kling

Kling 2.0 大师版针对专业创作场景优化，为高要求项目提供电影级输出质量。

wan2.6-r2v-flash

Alibaba

From Alibaba, supports Video and more—call it via the standard API.

wan2.6-r2v

Alibaba

From Alibaba, supports Video and more—call it via the standard API.

qwen3.6-flash

Qwen

Qwen 3.6 lightweight tier with text, image, and video input. Low latency and low cost, well-suited for high-volume classification, extraction, summarization, and simple agent workflows.

kimi-k2-instruct

Moonshot AI

Instruction-tuned for agent workflows with reliable tool calling and multilingual coding.

kimi-k2-thinking

Moonshot AI

Open-source thinking model built for long-horizon reasoning and hundreds of sequential tool calls across autonomous research, coding, and agent workflows.

qwen3.6-plus

Qwen

Enhanced Qwen model with strong bilingual (Chinese/English) comprehension, excelling at long-document analysis and structured output.

mimo-v2-pro

Xiaomi

Flagship agent model for complex workflows

minimax-m2.7

MiniMax

Multi-agent model for complex workflows

glm-4.7

Z.AI

Balanced model with strong Chinese understanding.

glm-4.6v

Z.AI

Multimodal model with strong image + text understanding.

minimax-m2.1

MiniMax

Cost-effective general model for balanced speed and quality.

glm-5-turbo

Z.AI

Designed for agent-based workflows such as OpenClaw.

kimi-k2.5

Moonshot AI

Top Chinese-English bilingual. Strong long context.

glm-5

Z.AI

Competitive general-purpose Chinese model.

glm-4.7-flash

Z.AI

Ultra-fast for Chinese. Very cheap.