China
Organisations et modèles d’IA de China
Organisations actives
27
Modèles suivis
570
Modèle le plus récent
oct. 2026
Dernières sorties de China
octobre 2026
Kimi K3.1 Pro is Moonshot AI's reasoning model with a 1M-token context window, image and video input, tool calling and structured output.
MiniMax M3.1 Flash is a multimodal coding model with a 1M-token context window, text, image and video input, and tunable reasoning depth.
DeepSeek V4.1 Flash 1017 is the 17 October 2026 update with improved reasoning, coding, agentic and visual performance, shorter task execution time and output length, native image understanding, a 1M-token context window and up to 384K output tokens.
septembre 2026
Ling 3.1 Flash is InclusionAI's hybrid reasoning language model with 560B total parameters and 25B activated parameters.
Ling 3.1 Flash is InclusionAI's hybrid reasoning language model with 560B total parameters and 25B activated parameters.
Meituan describes LongCat 2.5 Preview as a natively multimodal model for long-horizon tasks across terminals, browsers, GUIs, spreadsheets, and design tools. The announcement reports 1.6T total parameters (about 48B active) and a 1M-token context window.
Tencent Hunyuan image generation and reference-based editing through GMI Cloud, up to 2K output.
MiMo V2.6 Flash is Xiaomi's fast, cost-efficient multimodal model on the MiMo Open Platform, accepting text, image, audio, and video input.
MiMo V2.6 Pro is Xiaomi's flagship multimodal model on the MiMo Open Platform, accepting text, image, audio, and video input. Passing service_tier fast selects its UltraSpeed route for up to 20× faster output.
Organisations de China
241 modèles
46 modèles
45 modèles
37 modèles
35 modèles
33 modèles
24 modèles
19 modèles
17 modèles
15 modèles
12 modèles
11 modèles
9 modèles
8 modèles
6 modèles
5 modèles
3 modèles
2 modèles
1 modèle
1 modèle
0 modèle
0 modèle
0 modèle
0 modèle
0 modèle
0 modèle
0 modèle
Modèles de China
Parcourez le catalogue complet, classé par organisation.
570 modèles
Alibaba
6 modèlesHappyHorse 1.1 is Alibaba's current audio-video generation model for text-to-video, first-frame image-to-video and reference-image video generation.
Wan 2.7 is a video generation model from Alibaba. It supports text-to-video, image-to-video with first and last frame control, and reference-to-video, where multiple reference images guide the style and content...
HappyHorse 1.0 is Alibaba's audio-video generation model for text-to-video, first-frame image-to-video, reference-image video generation and prompt-driven video editing.
Alibaba's video generation model, released in March 2026, supporting over 10 visual creation capabilities in a unified system. Wan 2.6 generates 1080p video at 24fps from text, images, reference videos, or audio,...
Reported upcoming Alibaba world model.
Reported upcoming Alibaba music generation model.
BAAI
3 modèlesBGE Multilingual Gemma2 is a multilingual text embedding model for retrieval across many languages.
BGE-M3 is a multilingual text embedding model supporting dense, sparse, and multi-vector retrieval over inputs up to 8,192 tokens.
A lightweight multilingual cross-encoder for rescoring text documents against a query.
Baidu
19 modèlesErnie 5.0 is a speech-to-text model from Baidu for transcription and spoken-language understanding workflows. It is suited to captioning, meeting notes, transcription, and speech analytics pipelines.
CoBuddy is a specialized code generation model developed by Baidu, engineered with targeted optimizations for Coding and AI Agent scenarios. It delivers exceptional performance, characterized by high inference throughput and ultra-low end-to-end (E2E) latency.
ernie-4.5-0.3b is an open-weight model currently exposed by ModelScope API-Inference.
ERNIE 5.1 is Baidu's current multimodal flagship model served through Qianfan.
PaddleOCR-VL 1.5 is a vision-language document parsing model served by SiliconFlow's China API.
A sophisticated text-based Mixture-of-Experts (MoE) model featuring 21B total parameters with 3B activated per token, delivering exceptional multimodal understanding and generation through heterogeneous MoE structures and modality-isolated routing.
ByteDance
37 modèlesSeedance 2.0 Mini is a video generation model from ByteDance. It supports text-to-video, image-to-video with first and last frame control, and multimodal reference-to-video with image, video, and audio inputs. It...
Seedance 2.5 is ByteDance's next-generation audiovisual video model for text-to-video, image-to-video, and reference-guided generation. It supports multimodal creative inputs, local video editing, multilingual creation, standard clips up to 30 seconds, and beta long-video generation up to 180 seconds.
Seedream 5.0 Pro is ByteDance's professional image generation and editing model for high-precision visual creation. BytePlus says it supports text-to-image, single image-to-image, and multi-reference image-to-image workflows with more precise control over positions and elements.
Seed 2.1 Turbo is ByteDance's multimodal model for coding and agent workflows, with text, image, and video input and a 256K context window.
Seedance 2.0 Mini (2026-06-15) is a lower-tier video generation model in ByteDance's Seedance 2.0 series. It supports multimodal video generation inputs such as text, image, and video.
Seed 2.0 Lite (2026-04-28) is a dated ByteDance release of the Seed 2.0 Lite multimodal model for text and visual inputs, aimed at lower-latency and cost-sensitive production use.
DeepSeek
46 modèlesDeepSeek V4.1 Flash 1017 is the 17 October 2026 update with improved reasoning, coding, agentic and visual performance, shorter task execution time and output length, native image understanding, a 1M-token context window and up to 384K output tokens.
DeepSeek V4.1 Flash is a 552B-backbone multimodal Mixture-of-Experts model with 8B active parameters during prefill and 16B during decode, a 1M-token context window, native image understanding, continuously controllable reasoning effort, and optimized agentic performance.
DeepSeek V4 Flash Vision Exp is DeepSeek's experimental multimodal vision-understanding model. It combines V4 Flash-level text, reasoning, and agent capabilities with image understanding.
DeepSeek V4 Pro 0813 is the 13 August 2026 version served by DeepSeek's stable deepseek-v4-pro API slug. It supports thinking and non-thinking modes, JSON output, tool calls, Chat Completions, Responses, Anthropic Messages, prefix completion, fill-in-the-middle completion, a 1M-token context window, and up to 384K output tokens. DeepSeek postponed discontinuation of the V4 Pro service to 12:00 Beijing Time on 14 September 2026; a future V4.1 Pro is intended to replace it.
DeepSeek V4 Flash 0731 is the official 31 July 2026 release that supersedes the V4 Flash preview, with substantially stronger agent capabilities and a DSpark speculative-decoding module. It supports native Responses API use, tool use, structured output, server-side web search, a 1M-token context window, and up to 384K output tokens.
DeepSeek V4 Flash is an efficiency-optimized Mixture-of-Experts model from DeepSeek with 284B total parameters and 13B activated parameters, supporting a 1M-token context window.
Dots Studio
2 modèlesDots3-Note Preview is an open-weight mixture-of-experts model from Dots Studio, with 16B active parameters out of 280B total. It is the lightest model in the Dots 3 family and is...
dots3-note Preview is Dots Studio's first open-weight dots3 model: a 280B-total-parameter, 16B-active multimodal Mixture-of-Experts model for reasoning, tool use, long-horizon agents, code, and multimodal understanding.
Inclusion AI
17 modèlesLing 3.1 Flash 1019 is InclusionAI's upgraded Ling 3.1 Flash, a hybrid reasoning mixture-of-experts model with 560B total parameters, 25B activated parameters and a 1M-token context window. Further post-training improves reasoning and real-world agentic performance.
Ling 3.1 Flash is InclusionAI's hybrid reasoning language model with 560B total parameters and 25B activated parameters.
Ling 3.1 Flash is InclusionAI's hybrid reasoning language model with 560B total parameters and 25B activated parameters.
Ling 3.0 Flash Sante is Inclusion AI's health and medicine enhanced 124B-parameter mixture-of-experts model with approximately 5.1B active parameters per token. It is designed for medical knowledge reasoning, clinical safety, evidence-based retrieval, and long-horizon medical tasks while retaining general reasoning, coding, and agentic capabilities.
Ling 3.0 Flash VL is Inclusion AI's 124B-parameter mixture-of-experts model with approximately 5.1B active parameters per token. It builds on Ling 3.0 Flash with native visual perception and advanced visual-agent capabilities while retaining strong language, reasoning, coding, and agentic performance.
Ling 3.0 Flash Fin is a finance-enhanced version of Inclusion AI's Ling 3.0 Flash, a 124B-parameter mixture-of-experts model with approximately 5.1B active parameters. It is optimized for complex, multi-step investment workflows and long-horizon planning while retaining general reasoning, coding, and mathematics capabilities.
Kling AI
11 modèlesKling v3.0 Pro is Kuaishou's premium video generation model, offering higher visual quality than the Standard tier. It supports text-to-video and image-to-video workflows, with first-frame and last-frame control for precise...
Kling v3.0 Standard is a video generation model from Kuaishou. It supports text-to-video and image-to-video workflows, with first-frame and last-frame control for guided scene composition. Clips range from 3 to...
Kling Video 3.0 native 4K text-to-video and image-to-video. Novita route staged pending pricing, deployment and live validation.
Kling 2.1 Master — premium text-to-video and image-to-video, 5–10s clips, 1080p, 24fps.
Kling 2.5 Turbo — fast text-to-video and image-to-video with first-last-frame mode, 5–10s clips, 1080p, 24fps.
Kling 2.6 — text-to-video and image-to-video with motion control + audio control (pro mode), variable duration, 1080p, 24fps.
KwaiKAT
8 modèlesKAT-Coder-Pro V2.5 is a high-end coding agent model in KwaiKAT's KAT-Coder series, built for complex software engineering, repository-scale tasks, and autonomous development workflows.
Kat Coder Air V2.5 is a fast, lightweight coding model in KwaiKAT's KAT-Coder series, designed for interactive development, rapid code iteration, and efficient coding agents.
Keye VL 2.0 30B A3B is Kuaishou's multimodal reasoning model with image, video, tool-use, and long-context support.
Kat Coder Air V1 is a lightweight coding model in the KwaiKAT KAT-Coder series.
Kat Coder Exp 72b is the latest high-performance model in KwaiKAT's KAT-Coder series, designed for complex enterprise-grade software engineering and SaaS integration. This entry tracks a preview release and capabilities may change before general availability.
Kat Coder Pro was a high-performance model in KwaiKAT's KAT-Coder series, designed for complex enterprise-grade software engineering and SaaS integration.
MedAIBase
1 modèleantangelmed is an open-weight model currently exposed by ModelScope API-Inference.
Meituan
9 modèlesMeituan describes LongCat 2.5 Preview as a natively multimodal model for long-horizon tasks across terminals, browsers, GUIs, spreadsheets, and design tools. The announcement reports 1.6T total parameters (about 48B active) and a 1M-token context window.
LongCat 2.0 is Meituan's current LongCat generation model, replacing the earlier LongCat 2.0 Preview provider model ID for general text generation.
LongCat 2.0 Preview was Meituan's beta LongCat generation model and was replaced by the stable LongCat 2.0 provider model ID on June 29, 2026.
LongCat Flash Chat was a legacy language model from Meituan for chat, writing, analysis, and general-purpose automation. Meituan retired this legacy model on May 29, 2026 as part of its migration toward LongCat 2.0 Preview.
LongCat Flash Chat 2602 Exp was an experimental legacy Meituan LongCat chat model snapshot. Meituan retired this legacy model on May 29, 2026 as part of its migration toward LongCat 2.0 Preview.
LongCat Flash Lite was a legacy Meituan LongCat model. Meituan retired this legacy model on May 29, 2026 as part of its migration toward LongCat 2.0 Preview.
MiniMax
45 modèlesMiniMax M3.1 Flash is a multimodal coding model with a 1M-token context window, text, image and video input, and tunable reasoning depth.
MiniMax H3 is MiniMax's next-generation general-purpose multimodal video model for text-to-video, first/last-frame image-to-video, multimodal reference generation and video editing.
Music 3.0 is MiniMax's released music-generation model for producing complete vocal or instrumental tracks from natural-language prompts and optional lyrics. It improves creative-intent adherence, arrangement, mixing, and vocal expression.
Music 3.0 is MiniMax's released music-generation model for producing complete vocal or instrumental tracks from natural-language prompts and optional lyrics. It improves creative-intent adherence, arrangement, mixing, and vocal expression.
MiniMax M3 is MiniMax's frontier multimodal coding and agentic model. It supports up to 1M of input context for long documents, codebases, and multi-step workflows across text, image, and video inputs.
MiniMax M3 is MiniMax's frontier multimodal coding and agentic model. It supports up to 1M of input context for long documents, codebases, and multi-step workflows across text, image, and video inputs.
Moonshot
24 modèlesKimi K3.1 Pro is Moonshot AI's reasoning model with a 1M-token context window, image and video input, tool calling and structured output.
Kimi K3 is Moonshot AI's 2.8-trillion-parameter native multimodal model for software engineering, knowledge work, and deep reasoning. It uses Kimi Delta Attention, Attention Residuals, and a Stable LatentMoE architecture that activates 16 of 896 experts, accepts text, image, and video input, and supports a 1,048,576-token context window.
Kimi K2.7 Code is Moonshot AI's strongest coding model, built for long-context software engineering and agent workflows with native text, image, and video input plus always-on long thinking.
Kimi K2.6 is Moonshot AI's next-generation multimodal model, designed for long-horizon coding, coding-driven UI/UX generation, and multi-agent orchestration.
Kimi K2.6 Code Preview is a coding model from Moonshot for code generation, edits, and software tasks. This entry tracks a preview release and capabilities may change before general availability.
Kimi K2.5 is Moonshot AI's native multimodal model, delivering state-of-the-art visual coding capability and a self-directed agent swarm paradigm. It supports turning still images into animated clips as well as prompt-driven video workflows.
Nex AGI
5 modèlesNex N2.5 Max builds on DeepSeek-V4-Pro-Base, a text-only 1.6T-parameter sparse Mixture-of-Experts model with 49B active parameters, with Nex AGI's full-scale agent post-training for complex agent work, scientific modeling, engineering design, and complex frontend development.
Nex N2.5 Mini builds on Qwen3.5-35B-A3B-Base, an efficient hybrid vision-language sparse Mixture-of-Experts model, with Nex AGI's agent-focused post-training for high-speed instruction following, real-time tool execution, computer use, web browsing, and cost-effective deployment.
Nex N2.5 Pro builds on Qwen3.5-397B-A17B, a large hybrid vision-language sparse Mixture-of-Experts model, with Nex AGI's agent-focused post-training for complex reasoning, multi-agent orchestration, advanced software engineering, computer use, and web tasks.
Nex N2 Mini is an open-weight agentic multimodal model tuned for coding, tool use, and long-horizon real-world tasks in a smaller 35B-class footprint.
Nex N2 Pro is an open-weight agentic multimodal model aimed at coding, tool use, terminal execution, and long-horizon real-world tasks in a 397B-class configuration.
OpenCompass
1 modèlecompassjudger-1-32b-instruct is an open-weight model currently exposed by ModelScope API-Inference.
Qwen
241 modèlesQwen Image 2.1 is an open-weight model for text-to-image generation and image editing. Its visual generation component has 7B parameters, with native transparent RGBA generation and editing, local edit controls, and support for up to 10 reference images.
Qwen's multimodal understanding model accepting text, image, audio and video input with text output, reasoning, tool calling and a million-token context window.
Qwen3.8-Max-0902 is a pinned Qwen Cloud snapshot of Qwen3.8-Max, further post-trained for coding and collaborative agent work. It improves engineering-scale coding, long-horizon autonomous development, multi-tool orchestration, enterprise tasks, scientific research, chart reasoning, document parsing and multimodal perception while retaining the 1M-token context window, thinking mode and built-in tool ecosystem. It is also available under the alias qwen3.8-max-2026-09-02.
Qwen 3.8 Flash is Qwen's fast multimodal model for reasoning, coding assistance, agentic workflows, visual understanding, and long-context applications. Model Studio advertises a million-token context window and OpenAI- and Anthropic-compatible APIs.
Wan 3.0 Video Prime is the speed-optimized Alibaba Cloud Model Studio Wan 3.0 video generation model. It keeps the standard model's all-in-one text-to-video, first/last-frame, multimodal reference, video editing, video extension, and native audio capabilities while improving end-to-end generation speed.
Qwen 3.8 27B is a dense 27.8-billion-parameter open-weight vision-language model from the Qwen Team at Alibaba. It supports text, image and video inputs, a 262,144-token native context window extendable to 1 million tokens, tool use and controllable reasoning.
StepFun
35 modèlesStep 5 Preview is StepFun's flagship multimodal reasoning model for software engineering, agentic tasks and professional knowledge work. Its sparse MoE architecture has 600 billion total parameters and 27 billion active parameters per token, with a 1 million-token context window and text, image and video input.
Step 3.7 Flash is StepFun's flagship multimodal reasoning model for real-world agent, coding, and workflow tasks. It supports text, image, and video inputs with low-latency tool use.
Audio and text conversation model with text output and paralinguistic understanding.
Preview audio and text conversation model with reasoning, function calling and JSON output.
Tencent Hunyuan
12 modèlesTencent Hunyuan image generation and reference-based editing through GMI Cloud, up to 2K output.
Hy4 Preview is Tencent Hunyuan's next-generation Mixture-of-Experts model for coding agents, complex tool-use workflows, and productivity tasks. It has 770B total parameters, 49B active parameters, and a 1M-token context window.
Hy-MT2-1.8B is a compact 1.8B-parameter translation model from Tencent. It supports 33 language pairs and five Chinese dialect and minority-language pairs, with workflows for structured, delimiter-based, contextual, glossary-based, and style-guided...
Hy-MT2-30B-A3B is Tencent's flagship translation model in the Hy-MT2 family. It supports 33 language pairs and five Chinese dialect and minority-language pairs, with workflows for structured, delimiter-based, contextual, glossary-based, and...
Hy-MT2-7B is a 7B-parameter translation model from Tencent. It supports 33 language pairs and five Chinese dialect and minority-language pairs, with workflows for structured, delimiter-based, contextual, glossary-based, and style-guided translation.
Hy3 is Tencent Hunyuan's high-efficiency Mixture-of-Experts model for agentic workflows, reasoning, coding, and long-context text generation.
Xiaomi
15 modèlesMiMo V2.6 Flash is Xiaomi's fast, cost-efficient multimodal model on the MiMo Open Platform, accepting text, image, audio, and video input.
MiMo V2.6 Pro is Xiaomi's flagship multimodal model on the MiMo Open Platform, accepting text, image, audio, and video input. Passing service_tier fast selects its UltraSpeed route for up to 20× faster output.
MiMo V2.5 ASR is Xiaomi's bilingual Chinese-English speech-recognition model with automatic code switching, Chinese dialect and lyrics transcription, and support for noisy, far-field, and multi-speaker audio.
MiMo V2.5 TTS VoiceClone replicates a voice from a short reference-audio sample and synthesizes speech with the cloned timbre, including cross-language use and style control.
MiMo V2.5 TTS VoiceDesign creates a custom synthetic voice from a natural-language description without reference audio, with control over accent, age, timbre, pace, and recording texture.
MiMo V2.5 TTS VoiceClone replicates a voice from a short reference-audio sample and synthesizes speech with the cloned timbre, including cross-language use and style control.
z.AI
33 modèlesGLM 5.3 Flash is Z.AI's 320B-parameter multimodal coding and agentic model for software engineering, sustained agent workflows, and complex reasoning, with 18B active parameters and a 1M-token context window. It accepts text, image, and video inputs, supports tool calling and structured outputs, and uses mandatory reasoning with low, high, and max effort levels.
GLM 5.3 is Z.AI's frontier coding and agentic model, with low, high, and max reasoning effort levels and a 1M-token context window on the GLM Coding Plan.
GLM-5.2 introduces a robust 1M-token context and advanced, multi-effort coding capabilities to significantly enhance performance on long-horizon tasks. Its new IndexShare architecture and improved MTP layer simultaneously boost efficiency by reducing per-token FLOPs and increasing speculative decoding lengths. A 743B-parameter model in Zhipu AI's GLM series, designed to plan, execute, and iterate autonomously on extended, engineering-grade tasks.
GLM 5.2 is Z.AI's open-weights model for coding and long-horizon agentic tasks, with a 1M context window and high/max reasoning effort controls.
GLM-5.1 delivers a major leap in coding capability, with particularly significant gains in handling long-horizon tasks.
GLM 5V Turbo is Z.ai's first native multimodal agent foundation model, built for vision-based coding and agent-driven tasks. It supports turning still images into animated clips as well as prompt-driven video workflows.