AIUpdateWatch Intelligence
The main Chinese model families are treated as first-class entries in AIUpdateWatch model, benchmark, pricing, open-model and source-health coverage.
Daily registry
Capabilities, access, context and prices are shown only where a provider or reliable primary source publishes them. Chinese yuan values are never relabeled as US dollars.
Best fit: Long-context reasoning, coding and agent workflows
Modalities: Text
Context: 1M
Availability: DeepSeek API, app and web
Caveat: Official August 13 production revision DeepSeek-V4-Pro-0813. Prices shown are the current pre-August 17 tariff; announced peak/off-peak rates are future-dated. No independent 0813 benchmark package was verified by the edition cutoff.
Open official source (opens in a new tab)Best fit: Large-context efficiency and open-model ecosystem
Modalities: Text
Context: 1M
Availability: DeepSeek API through the explicit deepseek-v4-flash and deepseek-v4-pro model names; legacy aliases are past their retirement deadline.
Caveat: DeepSeek primary documentation identifies April 24 as the V4 Preview API availability date. Later third-party score snapshots vary and should be read with their evaluation date and methodology.
Open official source (opens in a new tab)Best fit: Complex multi-step reasoning and coding in the Qwen ecosystem
Modalities: Text, reasoning, coding, tools and structured output
Context: See current Model Studio catalog
Availability: Alibaba Cloud Model Studio; regional endpoints vary
Caveat: Pricing, endpoint names and availability differ between China and international Model Studio regions.
Open official source (opens in a new tab)Best fit: Long-context agentic work and software engineering
Modalities: Text, software engineering, knowledge work, deep reasoning and tool calling
Context: 1M tokens
Availability: Kimi API platform; region and account eligibility vary
Caveat: Prices are official CNY rates per million tokens and must not be displayed as USD.
Open official source (opens in a new tab)Best fit: Long-horizon coding and autonomous agent workflows
Modalities: Text, reasoning, coding, long-running agents and tools
Context: 1M tokens
Availability: Zhipu BigModel platform; regional access varies
Caveat: Provider capability statements require independent task-specific reproduction.
Open official source (opens in a new tab)Best fit: Unified multimodal tasks and Chinese-language applications
Modalities: Unified text, image, video and audio understanding/generation
Context: 128K tokens
Availability: Baidu Qianfan and ERNIE services; international catalog availability varies
Caveat: The displayed API price is from Baidu’s international Qianfan catalog; China-region billing differs.
Open official source (opens in a new tab)Best fit: Production-oriented agent, coding and multimodal tasks
Modalities: General agents, coding and multimodal workflows
Context: See current Volcengine Ark catalog
Availability: Volcengine Ark; regional availability varies
Caveat: Volcengine pricing uses tiered regional tables; verify the exact model ID and input-length tier.
Open official source (opens in a new tab)Best fit: Agentic reasoning, coding and long-context work
Modalities: Text, multimodal chat input, coding, tool use and long-context agents
Context: 1M tokens
Availability: MiniMax API platform and compatible endpoints
Caveat: Use the current regional pricing table and distinguish API pay-as-you-go from Token Plan subscriptions.
Open official source (opens in a new tab)Best fit: Low-cost multimodal reasoning and computer-use style tasks
Modalities: Multimodal reasoning, visual interaction, coding and agent workflows
Context: See current StepFun model catalog
Availability: StepFun open platform; regional access varies
Caveat: Official prices are CNY per million tokens. Confirm current quotas and model lifecycle notices.
Open official source (opens in a new tab)Best fit: Efficient mixture-of-experts reasoning in the Tencent ecosystem
Modalities: Text, hybrid reasoning, math, science, long documents and agents
Context: 224K max input; 32K max output
Availability: Tencent Cloud Hunyuan / migration path to TokenHub
Caveat: Tencent is migrating newer model access toward TokenHub; verify the active platform before integration.
Open official source (opens in a new tab)Best fit: Dense layouts, small text, multilingual rendering, interfaces and infographics
Modalities: Text-to-image generation and image editing
Context: Up to 4.5K-token image instructions
Availability: Qwen and Alibaba Cloud ecosystem; verify regional endpoint availability
Caveat: Capabilities and examples are provider-published; independent benchmark reproduction is pending.
Open official source (opens in a new tab)Best fit: Real-time Flash variant and higher-fidelity Plus variant
Modalities: Text-to-speech, voice cloning, style control and non-verbal tags
Context: 16-language text-to-speech release
Availability: Alibaba Cloud Model Studio; rollout and regional availability may vary
Caveat: The 300ms-level first-packet latency and multilingual results are provider-reported; verify independently for production workloads.
Open official source (opens in a new tab)| Provider | Tracked model | Published context | Primary strength | Source |
|---|---|---|---|---|
| DeepSeek | V4 Flash | 1M tokens | Large-context efficiency and open-model ecosystem | Official source (opens in a new tab) |
| Alibaba Cloud | Qwen3.7-Max | See current Model Studio catalog | Complex multi-step reasoning and coding in the Qwen ecosystem | Official source (opens in a new tab) |
| Moonshot AI | Kimi K3 | 1M tokens | Long-context agentic work and software engineering | Official source (opens in a new tab) |
| Zhipu AI | GLM-5.2 | 1M tokens | Long-horizon coding and autonomous agent workflows | Official source (opens in a new tab) |
| Baidu | ERNIE 5.0 | 128K tokens | Unified multimodal tasks and Chinese-language applications | Official source (opens in a new tab) |
| ByteDance | Doubao Seed 2.1 | See current Volcengine Ark catalog | Production-oriented agent, coding and multimodal tasks | Official source (opens in a new tab) |
| MiniMax | MiniMax-M3 | 1M tokens | Agentic reasoning, coding and long-context work | Official source (opens in a new tab) |
| StepFun | Step 3.7 Flash | See current StepFun model catalog | Low-cost multimodal reasoning and computer-use style tasks | Official source (opens in a new tab) |
| Tencent | Hunyuan A13B | 224K max input; 32K max output | Efficient mixture-of-experts reasoning in the Tencent ecosystem | Official source (opens in a new tab) |
Chinese and non-Chinese models use the same benchmark-version, configuration and source-quality requirements.
Pricing remains in its published currency and region. Conversion, when shown, must include a timestamp and methodology.
Regional availability, documentation language, licensing and export-control uncertainty are kept beside the comparison.