To try a named Chinese AI model from an OpenAI-compatible app, choose the actual hosting provider as well as the model. A free chat app or downloadable model does not automatically include a free hosted API. The examples below describe provider offers checked on 30 October 2026; account eligibility and live availability were not tested.
DeepSeek V4.1 Flash: NVIDIA prototype endpoint
NVIDIA's model catalog advertises a free API endpoint for prototyping. Its example uses an NVIDIA API key, OpenAI-compatible base URL https://integrate.api.nvidia.com/v1 and model ID deepseek-ai/deepseek-v4.1-flash. Developer access is for development, testing, research and evaluation, not production. Check your account eligibility and current model limits; a fixed free request allowance was not established.
This is NVIDIA-hosted access, not the direct DeepSeek account or API key. DeepSeek's own API is metered. NVIDIA model page and example · NVIDIA developer access · DeepSeek direct API and other routes
Qwen3.8-27B: Groq free plan
Groq's free-plan rate-limit table lists qwen/qwen3.8-27b. Groq's OpenAI-compatible base URL is https://api.groq.com/openai/v1; use a Groq account and key. Limits apply at the organization level and depend on the model and plan. Check the current table and your account before use.
Groq free-plan limits · Groq OpenAI compatibility · More Qwen access options
QwenCloud: time-limited new-user quota
QwenCloud documents model-specific free quota for eligible new users in the Singapore region, typically valid for 90 days. Some models are excluded; new users must complete payment information to activate the quota. After it is exhausted, usage can be billed unless Free quota only is enabled. Check the model and expiry in your own account before calling it.
QwenCloud quota and stop setting · Account and endpoint details
GLM-4.7-Flash: free price, availability to verify
Z.ai's international price table lists GLM-4.7-Flash input and output as Free. Its current Chat Completion reference does not list that model among its available options, so the price table alone does not establish that a particular account can select it. Check your account's current model list and limits before using it. Z.ai lists Web Search separately at a per-use price; free model tokens do not make every tool free.
Z.ai's standard OpenAI-compatible base is https://api.z.ai/api/paas/v4 and uses a standard API key. Its Coding Plan is a different product. The mainland BigModel documentation shows glm-4.7-flash with a different endpoint; do not mix regional hosts and keys. We have not tested either account or model route.
Z.ai price table · Z.ai current model reference · BigModel mainland example · Standard API versus Coding Plan
GLM-4.7-Flash on Cloudflare Workers AI
Cloudflare lists @cf/zai-org/glm-4.7-flash as available on the Workers Free plan. For its OpenAI-compatible API, use your Cloudflare account ID in https://api.cloudflare.com/client/v4/accounts/{account_id}/ai/v1 and a Cloudflare Workers AI token. This is a Cloudflare-hosted route with its own account and credentials, separate from Z.ai.
Workers AI includes 10,000 Neurons per day shared across the account; the allocation resets at 00:00 UTC. Free-plan requests stop after the allowance is exhausted; usage beyond it on the Paid plan is billed. Check the model and account limits before use; this route has not been tested with an account or model request.
Cloudflare model and API · OpenAI-compatible endpoint · Workers Free model availability · Daily allocation and billing
Kimi and MiniMax on FreeTheAI
FreeTheAI's provider guide, updated 30 September 2026, lists Kimi K3 as fta/kimi/k3 and MiniMax M3 as fta/mmx/minimax-m3. It describes a verified-email account, a FreeTheAI API key and a daily check-in, with 50 free requests per account per day and one concurrent request. The OpenAI-compatible base is https://api.freetheai.org/v1.
These are FreeTheAI-hosted routes, not official Kimi or MiniMax keys and not ChinaGenAI's service. Its plans page separates hosted-model allowances from wallet-billed marketplace models. Model listings, email eligibility, regional access and actual inference may change; check the provider's current model page and your account before use. No account or model request was tested here.
Qiniu Conatus: time-limited composite model
Qiniu's Conatus promotion offers the anonymous composite model qiniu/conatus-free through an OpenAI-compatible API. Its official page says verified accounts not restricted by risk controls can receive 100 million combined input and output tokens per day, including cached input and shared by all keys on that account. Check the actual allowance in the provider console. Real-name verification is required; overseas eligibility and English-language support are unverified.
The offer runs through 30 November 2026, 23:59 Beijing time. The daily allowance resets at midnight Beijing time, and unused quota does not roll over; calls stop when it is exhausted and after the promotion ends, with no automatic paid fallback for this model. Qiniu also advertises a larger daily allowance after a single credited top-up above RMB 100; that top-up is not required for the stated verified-account allowance.
Use a Qiniu API key and base URL https://api.qnaigc.com/v1. Qiniu describes this as an anonymous composite model; it does not guarantee a named DeepSeek, Qwen or GLM model. It is separate from ChinaGenAI's automatic route. Official promotion, requirements and example
Other model families and the site API
For direct provider access, check the product and plan rather than assuming that a third-party free route also applies: Kimi routes · MiniMax keys and regions.
ChinaGenAI also offers a separate free OpenAI-compatible endpoint using the model label free. It selects its upstream automatically and does not promise any named model. Check its access and limits or compare more providers.
This comparison does not verify a particular account, geographic availability, card requirement or live inference. Provider terms and model listings can change.
Free API access and limits
Check the site free API access and limits before trying to obtain a site key. Read the current status and request limits before using a site key. It is not a source of official provider keys.
ChinaGenAI is not affiliated with the providers named here. This release contains setup guidance and explicitly labeled fictional examples, not claimed live model benchmarks.