Site descriptions, notices, and reviews are shown in the language in which they were submitted.
专注提供国产开源大模型 API 服务,涵盖 DeepSeek、Qwen、GLM 等。 所有模型部署于自有机器、自有算力运行,核心算力采用 NVIDIA Blackwell Ultra 架构(B300 / GB300),企业级硬件保障稳定可靠。 机房位于新疆,依托当地低电力成本运营;不依赖任何第三方上游——无转发、无中间环节,因此不受第三方 API 波动、限流、停服影响,服务持续稳定在线。 自建算力 + 低电费 + 开源模型无 per-token 授权费,三重成本优势,价格公道实惠,性价比突出。
Price listing
GLM 5.2
3 public records are listed; open the full page to review details as needed.
Current public prices are shown by model and channel. Token prices use CNY per 1M tokens unless another billing unit is shown.
| Channel | Provider model | Current price | Price trend | Uptime | Fake % | Latency |
|---|---|---|---|---|---|---|
| beta | glm-5.2 | 输入 ¥1.6 / 输出 ¥5.61 / 缓存 ¥0.4007 / 写入 ¥0 | 0% | — | — | — |
| default | glm-5.2 | 输入 ¥1.6 / 输出 ¥5.61 / 缓存 ¥0.4007 / 写入 ¥0 | +100% | — | — | — |
| test | glm-5.2 | 输入 ¥1.6 / 输出 ¥5.61 / 缓存 ¥0.4007 / 写入 ¥0 | +100% | — | — | — |
Uptime is the share of recent probes where the service was available; a higher value generally indicates better stability.
Model consistency uses model pass results to show whether the endpoint behaves like the target model.
Pricing and latency come from public monitoring data and are intended for side-by-side gateway comparison.
Data is provided for reference. Third-party gateways may experience outages or balance loss, so test first and add only a small balance when needed.