Our take
GMICloud is a boutique inference host with a compact, curated catalogue focused on Chinese-model access. It discloses per-token pricing and measured throughput for each of its 13 offers, though several trust and geographic details remain unverified.
Use this for access to the Chinese-model ecosystem — Tencent, Xiaomi, MiniMax, Z-AI GLM and DeepSeek variants — when you want disclosed throughput numbers to guide selection. Pick it for budget-conscious DeepSeek V4 Flash inference. Skip it if you need verified SOC 2, a confirmed EU-region endpoint, or a zero-retention option; or if your workload requires Western open-weights models.
- Does not train on prompts — confirmed in its privacy policy.
- Tencent HY3 Preview has the cheapest input price among this provider's own offers.
- Tencent HY3 runs at 64 TPS, the fastest measured throughput this provider discloses.
- 13 offers span five Chinese labs: Tencent, Xiaomi, MiniMax, Z-AI and DeepSeek.
- Compliance and geographic transparency gaps: headquarters country, EU-region endpoint, SOC 2 and zero-retention option are all unverified in our data.
- Several offers show weak throughput relative to their price tier — Xiaomi MiMo V2.5 at 13 TPS is the slowest measured offer despite mid-range pricing.
- Thin catalogue depth versus open-weight generalists: 13 offers with no Western open-weights represented.
Privacy & data handling
- ✓
- Yes
- ✗
- No
- No answer on record
The colour says whether the answer favours you, not whether it is a yes — not training on your prompts earns a green cross, no zero-retention option earns a red one.