AI inference providers from China

LLM Providers from China: Leaderboard & Inference Benchmarks

Compare providers headquartered in China using observed inference speed, model availability, and published datacenter coverage.

Headquarters:: China
Providers:: 5
With speed data:: 2

Fastest LLM providers from China

Ranks are recalculated within China from each provider’s same-model normalized speed score. A provider needs at least three exact models measured on another provider; raw catalog speed remains visible for providers without a rank.

LLM Providers: Speed Leaderboard & Benchmarks
Position	Provider	Speed measurements	Data coverage
Rank
#1	Baidu Qianfan	Speed measurements 500-token response ↗ 14.90 s Response starts ↗ 1.01 s Output speed ↗ 36.0 tok/s	Data coverage Models offered 7 Speed measured 7 Shared comparisons 7 Rating basis Based on shared models
#2	StreamLake	Speed measurements 500-token response ↗ 18.69 s Response starts ↗ 1.75 s Output speed ↗ 30.5 tok/s	Data coverage Models offered 20 Speed measured 12 Shared comparisons 12 Rating basis Based on shared models
Without enough measurements · 3 providers
—	DeepSeek2 shared measured models; at least 3 required	Speed measurements 500-token response ↗ 9.56 s Response starts ↗ 1.11 s Output speed ↗ 60.0 tok/s	Data coverage Models offered 2 Speed measured 2 Shared comparisons 2 Rating basis Not enough data
—	Nex AGI1 shared measured model; at least 3 required	Speed measurements 500-token response ↗ 6.95 s Response starts ↗ 0.72 s Output speed ↗ 92.3 tok/s	Data coverage Models offered 2 Speed measured 2 Shared comparisons 1 Rating basis Not enough data
—	Xiaomi2 shared measured models; at least 3 required	Speed measurements 500-token response ↗ 13.85 s Response starts ↗ 2.36 s Output speed ↗ 43.5 tok/s	Data coverage Models offered 2 Speed measured 2 Shared comparisons 2 Rating basis Not enough data

See the complete methodology and worked rating example.

Inference market in China

How to compare LLM providers from China

This directory currently identifies 5 AI inference providers headquartered in China, including Baidu Qianfan, DeepSeek, Nex AGI, StreamLake. 2 providers have at least three shared model measurements and receive a local rank. The country view is useful for vendor discovery, but headquarters alone does not determine where inference requests are processed.

Current speed leader in China

Baidu Qianfan currently ranks first among qualifying China-headquartered providers. Its published catalog median estimates a 500-token response in 14.90 seconds across 7 measured models. Rankings change as OpenRouter publishes new observations.

Headquarters is not data residency

A company based in China may operate infrastructure in several countries or rely on external hosting partners. Treat the headquarters filter as company-location information, then verify endpoint regions, privacy terms, subprocessors, and contractual residency guarantees before handling regulated or sensitive data.

Review published hosting regions

Providers on this page collectively publish 2 datacenter regions: SG, NL. Region availability can vary by model and route, so confirm the current deployment location with the selected provider.