Heuristic signals on whether an OpenAI-/Anthropic-compatible endpoint serves the model it claims — catch model-swapping, quantization & silent context truncation. Zero-dependency Python CLI. Signals, not proof.
python cli benchmark proxy tokenizer verification openai gpt claude honesty llm anthropic ai-gateway deepseek llm-gateway openai-compatible model-integrity api-relay model-swapping downgrade-detection
-
Updated
Sep 4, 2026 - Python