AI
AIHubMix
aihubmix.com
5.0/ 10
One of the fastest-updating relay stations with the newest models — GPT-5.5, Claude Opus 4.7, Gemini 3.1 Pro. Active …
SECURITYAAA
UPTIME99.5%
MODELS10
PAYMENTUSD
Independent, developer-run reviews of every major relay station serving GPT-4o, Claude Sonnet, DeepSeek V3 and 40+ other models from inside China. No affiliate-only puff pieces.
Real workloads, real prices, real failure modes. Each has been re-tested in the last 14 days.
One of the fastest-updating relay stations with the newest models — GPT-5.5, Claude Opus 4.7, Gemini 3.1 Pro. Active …
Global relay with 24 CN2 nodes worldwide — supports GPT-4, Claude 3, Gemini 2.5. Excellent for international users with …
The international multi-model gateway. Full model coverage with the newest releases, but charges official price + markup …
Each review reruns our six-axis benchmark on the live API — not the marketing page.
USD aggregator routing vs. a CNY-native veteran. Cheaper at scale, but depends on whether your finance team will accept USD invoices.
If you only call DeepSeek V3, you're paying a relay tax. If you mix Claude + GPT, an aggregator pays for itself.
No paid placement. No vendor-supplied data. Each platform runs through the identical pipeline before it shows up on this site.
Surface candidate from forums, Github, WeChat groups.
Sign up under our test identity. Note payment methods, KYC asks.
TLS, key rotation, payment processor, data residency.
p50 / p99 latency, sustained TPS, cold-call variance.
Run a fixed 500-call agentic test across 6 models.
Effective $/1M tokens after surcharges & discount codes.
Score, disclose any affiliate link, re-score every 30 days.
Lowest effective $/1M tokens after all surcharges and minimum top-ups.
Audited TLS, rotatable keys, no client logs, transparent payment trail.
p50 under 250ms on GPT-4o calls — tested from Shanghai, Beijing, Shenzhen.
100+ models incl. Claude Sonnet 4, Gemini 2.5, Llama 4, DeepSeek V3.