ALL SYSTEMS OPERATIONALlast_scan 2026-11-01 16:42 SGTprobes 176 relays × 8 providersnext_scan in …
rankings/302-ai
3

302.AI

CN
riskyscore 5.4 — latency 1458ms and 96.7% uptime dragged this below the 6.0 risky floorlast probed 2026-11-01
Quality score
5.4
Get this relay →
Latency
1458 ms
Uptime (30d)
96.7%
Region
CN
Payment
wechat

Supported models

claudegptgeminianthropicopenaideepseekqwenmistral

About 302.AI

302.AI is an AI API relay serving the China region, advertising access to Anthropic (Claude), OpenAI (GPT), Google Gemini, DeepSeek, Alibaba Qwen, and Mistral. Based on our latest probe (2026-11-01) the endpoint responded in 1458 ms with 96.7% 30-day uptime, giving it a composite Quality Score of 5.4. It sits in the Risky band — either the composite score is below 6.0 or the source data flagged it for manual review in our ranking. Accepted payment methods on their site include wechat. Our authenticity probe uses model-specific fingerprint prompts to detect distillation, silent model substitution, or response caching — probe-based verification is a placeholder in the current MVP and will replace this paragraph as the pipeline lands.

Baseline model pricing

upstream $ per 1M tokens · models.dev

What the underlying model providers charge. Compare against what the relay quotes on their site to spot markup.

FamilyModelInputOutputContext
Anthropic (Claude)Claude Haiku 5.5$0.100$0.5001M
Claude Sonnet 5.5$2.00$10.001M
OpenAI (GPT)GPT-6.1 Terra$1.00$5.001.1M
GPT-6.1 Sol$2.00$10.001.1M
Google GeminiGemini 4 Argon$2.00$10.001.0M
Gemini 3.8 Flash$0.750$3.751.0M
DeepSeekDeepSeek V4 Flash$0.150$0.6001.0M
DeepSeek V4 Flash Vision Exp$0.150$0.6001.0M
Alibaba QwenQwen3.8 Omni Flash$0.150$0.4701M
Qwen3.8 Flash$0.150$0.4701M
MistralDeepSeek V4.1 Flash 1017$0.300$1.201.0M
Mistral Large 4.0.8$0.700$2.101.0M

Model authenticity analysis

MVP placeholder

Our authenticity probes are being built. Once live, this section will show whether 302.AI’s claimed models pass response-fingerprint checks against the upstream provider. See Methodology.