These are compact, highly efficient versions of OpenAI's GPT-5.4 model, optimised for speed and cost rather than maximum capability.
OpenAI's own Codex platform demonstrates the intended use: GPT-5.4 handles planning and coordination while GPT-5.4 mini subagents work in parallel on narrower tasks like searching a codebase or reviewing files.
The launch follows OpenAI's release of GPT-5.4 earlier this month, which introduced mid-response course correction, improved deep web research, and enhanced long-context reasoning.
In Codex, it uses only 30 percent of the GPT-5.4 quota.
Ключевые утверждения в источнике B
these are the company's 'most capable small models yet.' In ChatGPT, GPT-5.4 mini is available to Free and Go users via the 'Thinking' option found in the plus menu.
these are the company’s ‘most capable small models yet.’ The company says the new models bring many of the capabilities of the larger GPT 5.4 model but in a lighter format.
the model can also run more than twice as fast while still achieving performance close to the larger GPT 5.4 model in certain tests.
It uses only 30 per cent of the GPT 5.4 quota, allowing developers to handle simpler coding tasks at roughly one-third of the cost.
Текстовые доказательства
Доказательства из источника A
ключевое утверждение
These are compact, highly efficient versions of OpenAI's GPT-5.4 model, optimised for speed and cost rather than maximum capability.
Ключевое утверждение, на котором строится подача материала.
ключевое утверждение
In Codex, it uses only 30 percent of the GPT-5.4 quota.
Ключевое утверждение, на котором строится подача материала.
Доказательства из источника B
ключевое утверждение
According to OpenAI, these are the company's 'most capable small models yet.' In ChatGPT, GPT-5.4 mini is available to Free and Go users via the 'Thinking' option found in the plus menu.
Ключевое утверждение, на котором строится подача материала.
ключевое утверждение
According to OpenAI, these are the company’s ‘most capable small models yet.’ The company says the new models bring many of the capabilities of the larger GPT 5.4 model but in a lighter for…
Ключевое утверждение, на котором строится подача материала.
выборочный акцент
It uses only 30 per cent of the GPT 5.4 quota, allowing developers to handle simpler coding tasks at roughly one-third of the cost.
Возможное выборочное акцентирование отдельных аспектов истории.
Доказательства паттернов искажения
Источник B · Эффект фрейминга
It uses only 30 per cent of the GPT 5.4 quota, allowing developers to handle simpler coding tasks at roughly one-third of the cost.