2026-Q2 · Tested 2026-05-15
Is it really Claude?
We run a deterministic fingerprint against the official Anthropic API at temperature 0. If a gateway's Sonnet doesn't match, we say so — publicly.
Methodology v2 (mock data). 20 prompts · deterministic match · tokenizer + refusal + cutoff corroboration. Full design in methodology. Prompt text is not published (anti-gaming).
SUSPECT3 gateways
SUSPICIOUS3 gateways
LIKELY11 gateways
API.chat
Claude Sonnet 4.6 · match 9/20 · score 55
Apixo
Claude Sonnet 4.6 · match 9/20 · score 58
AIMLAPI
Claude Sonnet 4.6 · match 10/20 · score 60
OpusClaw
Claude Sonnet 4.6 · match 10/20 · score 62
IzziAPI
Claude Sonnet 4.6 · match 10/20 · score 65
Codex Maximum
Claude Sonnet 4.6 · match 11/20 · score 68
DMXAPI
Claude Sonnet 4.6 · match 11/20 · score 68
ClawHive
Claude Sonnet 4.6 · match 12/20 · score 70
GetGoAPI
Claude Sonnet 4.6 · match 12/20 · score 72
Wellflow
Claude Sonnet 4.6 · match 16/20 · score 82
Quatarly
Claude Sonnet 4.6 · match 16/20 · score 86
AUTHENTIC3 gateways