AIBench.cc
Claude API

Claude API Check | Anthropic Latency, Prompt Cache and Model Purity

Check Claude API and Anthropic-compatible channels: API latency, TTFT, prompt cache read/write, model purity, rate limits and real token cost.

Quick Answer

A Claude API check verifies Anthropic protocol behavior and reads cache_read_input_tokens, cache_creation_input_tokens, rate limit headers, model identity and TTFT to judge whether a Claude channel is stable and truly supports prompt cache.

How It Works

  1. 1Choose the native Anthropic Claude protocol.
  2. 2Enter Claude base URL, model id and a temporary API key.
  3. 3Run a long-prefix cache probe and compare cache creation with cache read.
  4. 4Record TTFT, P95 latency, rate limit headers and model identity signals.

Use Cases

  • Verify Claude Sonnet or Opus availability.
  • Check whether a relay preserves Anthropic prompt cache fields.
  • Compare Claude providers by first-token latency and rate limits.
  • Investigate Claude authenticity or downgrade risk.

Core Metrics

cache_read_input_tokens

Claude native prompt cache read tokens

cache_creation_input_tokens

Claude prompt cache creation tokens

anthropic-ratelimit

Anthropic rate limit response headers

Model identity

Consistency between model field and self-reported family

Claude prompt cache is a core signal

For long-context workloads, Claude prompt cache changes both cost and latency. If a gateway strips cache fields, you may pay full input cost without realizing it.

Common Claude relay risks

Risks include opaque model mapping, missing cache fields, hidden rate limit headers, incompatible error codes and abnormal model-family self-identification.

Related Search Terms

Claude API checkClaude API latency testAnthropic API checkClaude prompt cacheClaude relay test

FAQ

Can Claude API checks detect model swapping?

They can provide risk signals. AIBench combines model fields, self-identity, tokenizer signals and protocol behavior, but does not present a single probe as legal proof.

Does a cache miss always mean the channel is bad?

No. It can be caused by model, account, prefix length or gateway policy. AIBench shows the raw evidence so you can retest.

Related Guides and Comparisons

How to Check Claude API Prompt CacheOpenAI API vs Claude API: Check Differences

Start a Real API Check

Paste a temporary test key and run the same probes for latency, model purity, cache, rate limits and cost.

Start Check