OpenAI: GPT-4o-mini (2024-07-18)

GPT-4o-mini (2024-07-18) is a text generation model from OpenAI, released on 18 July 2024. It sits in OpenAI's small model line, positioned by the provider as its most advanced small model and the successor to GPT-4 Omni in that tier. The versioned model id pins the snapshot to 2024-07-18, so behaviour stays consistent across calls rather than shifting with a floating alias.

The model accepts text, images and files as input and returns text only. It has a context window of 128,000 tokens and can produce up to 16,384 output tokens in a single response. Its knowledge cutoff is 31 October 2023, so anything after that date has to come from the prompt, from files you supply, or from a search tool. Because image input is supported, you can pass screenshots, scanned pages or diagrams alongside text instructions and get a text answer back.

Those capabilities suit a range of everyday developer tasks. Document and file input works well for summarisation, extraction and question answering over contracts, reports or PDFs, while image input covers reading charts, receipts, UI screenshots and photographed text. The 128k context lets you send long transcripts, codebases or multi-document bundles without chunking them first. Tool calling and structured outputs make it a reasonable choice for agents, function routing and pipelines that need JSON shaped to a schema, and the seed, temperature and penalty parameters give you control over how deterministic or varied the output is.

You call it through api14's OpenAI-compatible API using the model id "openai/gpt-4o-mini-2024-07-18". Because the interface follows the OpenAI chat completions shape, existing OpenAI SDKs and clients work with a base URL change. The request parameters it supports are frequency_penalty, logit_bias, logprobs, max_tokens, prediction, presence_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_logprobs, top_p and web_search_options. That covers sampling controls, token limits, stop sequences, log probability inspection, JSON and schema-constrained output, tool definitions and forced tool selection, and web search options.

Two constraints are worth noting before you commit. Output is text only, so if you need generated images, audio or embeddings, this model is not the fit. The knowledge cutoff of 31 October 2023 means time-sensitive questions need retrieval, file input or web search options rather than relying on the model's own knowledge. And the 16,384 token output ceiling bounds how long a single response can be, which matters for long-form generation tasks.