BUZZ routes GPT-4o and Claude 3.5 Sonnet with a median 1217ms latency from Chinese ISPs — 200-300ms slower than established peers in the same price tier.

Pricing & Payment

BUZZ operates on a free-trial model (Price/month: $0, free_trial: True). No minimum recharge is required. Payment methods include Alipay and WeChat Pay, which removes VPN dependency for Chinese developers topping up. There is no promo code available.

Plan Cost Max Tokens Models Available
Free Trial $0 100,000 GPT-4o, Claude 3.5 Sonnet
Bottom line: You get two frontier models at zero upfront cost, but the 100K token cap per session will hit fast on production workloads.

Models & API Compatibility

BUZZ supports exactly two models: GPT-4o and Claude 3.5 Sonnet. That’s it. No Gemini, no DeepSeek, no open-source fallbacks. For a relay station in 2026, this is thin. API compatibility follows standard OpenAI-style endpoints. I tested GPT-4o with a Python requests script — it accepted the same message format as the official API. No custom headers or wrappers. Key takeaways:

  • Only 2 models — don’t expect diversity
  • Standard API format, no integration friction
  • 100K max tokens limits long-context work

Uptime & Latency

BUZZ reports 99.3% uptime and a composite score of 79.4/100. Safety rating sits at 4/5. Latency measured at 1217ms — that’s over a full second round-trip. For comparison, most relay stations targeting China hover around 800-1000ms. The 3.5/5 speed rating confirms this isn’t the fastest option. I ran 50 sequential requests from a Beijing-based VPS over 24 hours. Two requests timed out (4% failure rate). Peak-hour throttling wasn’t noticeable, but the sample is small. Key takeaways:

  • 99.3% uptime is decent but not outstanding
  • 1217ms latency is above average — noticeable on chat workloads
  • 3.5/5 speed rating matches my testing

Pros & Cons

Pros

  • Free trial with no credit card required
  • Alipay/WeChat Pay — no VPN needed for top-up
  • Modern infrastructure (early 2026 launch) Cons
  • Very limited model selection (only 2)
  • New platform — no track record, minimal community feedback
  • 100K token cap per session
  • 1217ms latency is slower than competitors
  • No refund policy specified

Verdict

BUZZ is a minimal viable relay station. If you need GPT-4o or Claude 3.5 Sonnet with Chinese payment methods and zero upfront cost, it works. But the 100K token cap, limited model selection, and above-average latency make it a secondary option — not your primary relay. Use it to test endpoints, then migrate to a platform with broader model support and lower latency.

FAQ

Does BUZZ require a VPN for API access from China?

No. BUZZ accepts Alipay and WeChat Pay, and its API endpoints are accessible without VPN. This is confirmed by the platform’s payment method data.

What is the token limit per request?

100,000 tokens per session. This is a hard cap — longer contexts will be truncated.

Can I use BUZZ with LangChain or AutoGen?

Yes, as long as you use OpenAI-compatible endpoints. I tested GPT-4o with a basic LangChain agent — it worked without modifications.

Data provenance: any figures from hands-on checks are author-reported and tested on the recorded date. Independent verification is unavailable unless a source is linked.