New to Rust? Grab our free Rust for Beginners eBook Get it free →
OpenAI GPT-4o Mini API Pricing, Limits, and Current Status
GPT-4o mini remains available through the OpenAI application programming interface (API), but it is now a legacy choice rather than OpenAI’s recommended small model for a new build. OpenAI’s current model catalog recommends GPT-5.6 Sol for complex work, GPT-5.6 Terra for a balance of capability and cost, and GPT-5.6 Luna for cost-sensitive, high-volume workloads.
Upgrade Your ChatGPT Skills to Pro
- How to Use ChatGPT: All Basics Covered
- How to Write the Perfect ChatGPT Prompt: A 2024 Guide
- 10 Amazing ChatGPT Prompt Templates You Should Know
- ChatGPT Prompts for Applying for a Job
- Best ChatGPT Prompts for Writing and Content Creation
GPT-4o mini is a legacy API option
OpenAI still lists gpt-4o-mini for focused tasks, fine-tuning, and distillation. Availability does not make it the current default: GPT-5.5 followed the GPT-4.1 generation, and the July 2026 catalog now places GPT-5.6 at the front of the API lineup.
ChatGPT access is separate from API access. GPT-4.1 mini was a historical replacement for GPT-4o mini in ChatGPT, not the current model recommendation. If you want the current consumer product, the ChatGPT guide covers the GPT-5 generation you can use now.
| Question | Current answer |
|---|---|
| Can you select GPT-4o mini in ChatGPT? | No. GPT-4.1 mini was an earlier replacement, while the current lineup has moved through GPT-5.5 to GPT-5.6. |
| Can you call gpt-4o-mini through the API? | Yes. OpenAI still documents the model and its API endpoints. |
| Is API use included with a ChatGPT subscription? | No. API usage has separate token billing. |
| Does the base model create images? | No. It accepts image input and returns text. |
| What should a new API project evaluate first? | GPT-5.6 Luna for cost-sensitive volume, Terra for a capability-cost balance, or Sol for complex work. |
Capabilities and limits that affect your build
The current OpenAI model page documents a 128,000-token context window, a 16,384-token maximum output, and an October 1, 2023 knowledge cutoff. Your prompt, image input, tool results, and conversation history all consume context.
| Capability | GPT-4o mini |
|---|---|
| Text input and output | Supported |
| Image input | Supported |
| Image output | Not supported |
| Audio and video | Not supported by the base model |
| Structured outputs | Supported |
| Function calling | Supported |
| Streaming | Supported |
| Fine-tuning | Supported |
Existing applications can still use the model for extraction, classification, translation, tagging, compact assistants, and image analysis when its low token price matters more than newer reasoning and tool use. The October 2023 cutoff means current facts need a supplied source, while a new application should benchmark GPT-5.6 Luna before committing to GPT-4o mini.
Image generation needs a separate model
GPT-4o mini can inspect an image and answer with text. A request to draw or edit an image must go to an OpenAI image model because GPT-4o mini does not generate DALL-E 3 images.
Web search is a tool, not built-in knowledge
The base model does not know events after its cutoff unless you provide them. OpenAI supports web search through API tooling and also documents a specialized GPT-4o mini Search Preview model, with search fees added to token charges.
Audio uses a separate GPT-4o mini variant
OpenAI lists separate Transcribe, text-to-speech (TTS), Audio Preview, and low-latency audio variants under the GPT-4o mini family. Their endpoints, context limits, output types, and prices differ from the base gpt-4o-mini model.
GPT-4o mini API pricing
OpenAI bills the base model per token. The model catalog checked on July 15, 2026 lists the following rates per one million text tokens.
| Token type | Price per 1 million tokens |
|---|---|
| Input | $0.15 |
| Cached input | $0.075 |
| Output | $0.60 |
At the published rates, a workload with 100,000 input tokens and 20,000 output tokens costs $0.027 before tool charges: $0.015 for input and $0.012 for output.
API usage requires separate billing and project limits because ChatGPT subscriptions do not cover it.
When keeping GPT-4o mini still makes sense
Keep gpt-4o-mini when an existing integration, fine-tune, latency target, or token budget still beats GPT-5.6 Luna on your evaluation set, accounting for the older cutoff and weaker capability.
- Use it for high-volume classification, extraction, or translation with structured output.
- Use it for image understanding when text output is enough.
- Keep it for an existing fine-tune that still meets your quality and latency targets.
- Evaluate GPT-5.6 Luna first for a new cost-sensitive build, then move to Terra or Sol when the workload needs more capability.
Test the same prompts against GPT-5.6 Luna and compare task accuracy, malformed structured responses, latency, tool-use requirements, and total output-token cost before keeping or migrating from GPT-4o mini.
GPT-4o mini questions
Is GPT-4o mini still available?
Yes, gpt-4o-mini remains documented for the OpenAI API. GPT-4.1 mini was an earlier ChatGPT replacement, while OpenAI’s current model lineup has moved through GPT-5.5 to GPT-5.6.
Is GPT-4o mini free?
API use is billed separately at published token rates. A ChatGPT plan does not include API usage, and GPT-4o mini is no longer a selectable ChatGPT model.
Can GPT-4o mini generate images?
The base model accepts image input and produces text output. Use an OpenAI image model when your application needs generated or edited images.
Can GPT-4o mini browse the web?
Current information requires a web search tool or retrieved source material. Tool calls can add separate charges beyond the model token cost.
A small evaluation set should decide whether gpt-4o-mini’s measured cost, speed, and output quality beat GPT-5.6 Luna for your workload, not an old replacement note or the input price alone.
