gpt-5-mini
A lightweight, fast-inference model released by OpenAI, targeting low-latency and low-cost use cases. It has fewer parameters than the flagship GPT series and is suitable for developers' everyday rapid-query scenarios.
Timeline (last 90 days)
In a test of 360 clean texts, Jev made only 1 misjudgment, gpt-5-mini made 1 misjudgment, Gemini 2.5 Flash-Lite and Gemma each made 2 misjudgments, gpt-oss-safeguard-20b made 4 misjudgments, and Llama Guard 4 made 26 misjudgments.
In the short name field test, Jev caught 21 of 27 verified terms in English, Spanish, Russian, Japanese, and Chinese, while gpt-5-mini caught 25; in the other 16 languages Jev caught only 15 of 80, while gpt-5-mini caught 53.
Jev's detection rate for disguised-spelling names was 13 of 25, while gpt-5-mini's was 25 of 25; adding leet characters or dots to clean names raised Jev's probability by an average of 0.16 (0.32 in English), while gpt-5-mini had zero false positives on 125 disguised clean names.
gpt-5-mini had zero false negatives on 184 abusive comments, but falsely flagged 57 acceptable items, most labeled as harassment.
In the experiment, ChatGPT used gpt-5-mini, Gemini used gemini-3.5-flash-lite, and Perplexity used pplx-fast — all lightweight, fast models from their respective providers.
All Facts (5)
In a test of 360 clean texts, Jev made only 1 misjudgment, gpt-5-mini made 1 misjudgment, Gemini 2.5 Flash-Lite and Gemma each made 2 misjudgments, gpt-oss-safeguard-20b made 4 misjudgments, and Llama Guard 4 made 26 misjudgments.
50%UnverifiedIn the short name field test, Jev caught 21 of 27 verified terms in English, Spanish, Russian, Japanese, and Chinese, while gpt-5-mini caught 25; in the other 16 languages Jev caught only 15 of 80, while gpt-5-mini caught 53.
50%UnverifiedJev's detection rate for disguised-spelling names was 13 of 25, while gpt-5-mini's was 25 of 25; adding leet characters or dots to clean names raised Jev's probability by an average of 0.16 (0.32 in English), while gpt-5-mini had zero false positives on 125 disguised clean names.
50%Unverifiedgpt-5-mini had zero false negatives on 184 abusive comments, but falsely flagged 57 acceptable items, most labeled as harassment.
50%UnverifiedIn the experiment, ChatGPT used gpt-5-mini, Gemini used gemini-3.5-flash-lite, and Perplexity used pplx-fast — all lightweight, fast models from their respective providers.
50%