A bilingual directory of AI products built by Chinese teams, focused on availability, English support, API access, pricing and alternatives.
50+
launch profiles
2
languages
API
tracked
Reference matrix
Availability, language, API, terms
The site is organized around practical decisions, not just product names. Pick a path based on what you need to compare.
Filter by availability, English UI, API and payment.
Start from video ads, LLM APIs or alternatives.
Use transparent scoring, not generic popularity.
See recent checks, guides and comparison updates.
See tools that already support Chinese models.
Start from Sora, Runway, ChatGPT or Cursor.
Compare DeepSeek, Qwen, Kimi and MiniMax models, products, pricing paths and implementation guides.
Scannable profiles built from the same evidence model.
Kuaishou
Kling has an English-facing site and a broader creative and API platform beyond text-to-video generation.
Best fit · Creators, studios and growth teams that want a globally accessible Chinese creative studio for video, image, sound, effects and API-backed generation.
Coverage · 100/100
Alibaba Cloud
FashionChameleon belongs in AI Video because it is specifically about real-time human-garment video customization, with clear relevance to e-commerce content creation and virtual try-on research.
Best fit · Researchers and e-commerce video teams tracking real-time virtual try-on, garment switching and human-centric video customization methods from Chinese labs.
Coverage · 100/100
ByteDance / Volcano Engine
OmniHuman 1.5 has its own BytePlus product, API, docs and pricing surface, so it should be tracked separately from Seedance and generic BytePlus video generation.
Best fit · Teams building avatar videos, talking-character content, AI music videos, ads and short-form digital-human workflows through BytePlus API access.
Coverage · 100/100
DeepSeek
The English docs now make DeepSeek's current API shape explicit: V4-Pro/V4-Flash, dual OpenAI/Anthropic compatibility, thinking mode, tool calls, caching and agent-tool integrations.
Best fit · Developers evaluating Chinese model APIs for coding agents, reasoning, tool use and OpenAI/Anthropic-compatible migration.
Coverage · 100/100
Product news, API changes, support notes and pricing signals that can affect selection.
2026-09-07
Qwen's current official surfaces now need to be read as two separate layers: Qwen3.8-Max is listed as the flagship model in Qwen Studio, while the August 26 Qwen release describes Qwen3.8-Flash-Next as an open architecture preview and Qwen3.8-Flash as the production hosted line. The release lists Qwen3.8-Flash at $0.16 per million input tokens and $0.47 per million output tokens, with API access still marked as coming soon. This directory keeps the preview, hosted model and open-weight variants separate instead of treating them as one product. · 15/15 sources verified, recently checked
2026-08-28
Qwen3.8-Flash-Next is now tracked as the experimental open architecture behind the production Qwen3.8-Flash line. The official model card documents 125B total parameters with 6B activated, a vision encoder, 262,144-token native context extendable to 1M, and qwen-community-1.0 weights. · 15/15 sources verified, recently checked
2026-08-28
Z.AI released GLM-5.3-Flash under MIT with first-party API and local deployment paths. The natively multimodal model uses 320B total and 18B active parameters, a 1M context configuration, hybrid sparse and linear attention, and low, high or max reasoning effort. Existing GLM selection and Kimi comparison content now distinguishes this variant from the original text-only GLM-5.3 launch. · 9/9 sources verified, recently checked
2026-08-28
The current individual Token Plan page lists Lite at $6/month with 2,500 credits per 7 days, Standard at $18 with 10,000 credits per 7 days and Pro at $68 with 40,000 credits per 7 days. The supported surface includes Qwen3.8-Max, GLM-5.2, DeepSeek V4 Pro and Wan2.7 Image plus Qwen Code, Cline, Claude Code, Cursor, OpenCode, Codex, Kilo CLI and OpenClaw. Checkout prices and eligibility remain subject to the live page. · 1/1 sources verified
2026-08-28
Kimi Code 0.39.1 fixes IME first-character loss, persists permission mode per session, improves attachments, prevents large-workspace startup from remaining on Connecting, keeps Skill instructions out of user-message history and increases the update timeout. · 7/7 sources verified, recently checked
The HDC 2026 announcement has moved into public release: Flash weights appeared on June 30 and Pro weights on July 30 in the official openPangu Hugging Face organization.
NewsChinese models take four of OpenRouter's top fiveMedia calculations based on OpenRouter data put DeepSeek V4 Flash, Tencent Hy3 Preview, MiniMax M3 and Xiaomi MiMo V2.5 in the top four for the week of June 1-7.
NewsSeedance 2.1 Late-June Report Remains UnconfirmedThe reported late-June launch window has passed. As of August 4, ByteDance Seed and BytePlus still do not list an official Seedance 2.1 product, so the quality and pricing claims remain unconfirmed.
View all articlesA source-backed comparison of the GLM-5.3 family and Kimi K3 across coding evidence, multimodality, context, weights, API access and deployment options, updated for GLM-5.3-Flash.
Decision pageMiniMax M3 vs DeepSeek V4A practical comparison of MiniMax M3 and DeepSeek V4 Pro or Flash across native multimodality, text reasoning, hosted cost, long output, open weights and deployment licensing.
Decision pageKling AI vs RunwayAn August 2026 buyer-oriented comparison between Kling's globally accessible Chinese video workflow and Runway's broader Real-World Intelligence creative suite.
View all comparisons