OpenAI API Pricing Calculator
Estimate your 2026 API costs across the full OpenAI model family. Calculate spend based on token usage for GPT-4o, GPT-4.1, o3, o4-mini, and GPT-4o mini.
API Usage Estimator
Project your monthly OpenAI spend
GPT-4o for multimodal; GPT-4.1 for coding; o-series for complex reasoning.
How many API calls do you expect per day?
Includes system prompt, user query, and context.
For o-series models, this includes internal reasoning tokens.
Projected API Costs
Monthly spend breakdown
Enter your expected requests, token usage, and model selection, then click Estimate to see your projected API costs.
OpenAI Models Overview
Capabilities and context windows for the 2026 OpenAI model lineup.
| Model | Context Window | Best For |
|---|---|---|
| GPT-4o | 128K tokens | Multimodal tasks, vision, audio, general-purpose chat |
| GPT-4o mini | 128K tokens | High-volume lightweight tasks, classification, summarisation |
| GPT-4.1 | 1M tokens | Coding, long-context analysis, agentic workflows, SWE-bench |
| GPT-4.1 mini | 1M tokens | Balanced coding and general tasks at lower cost |
| GPT-4.1 nano | 1M tokens | Ultra-fast edge tasks, on-device, latency-sensitive apps |
| o3 | 200K tokens | Complex reasoning, maths, science, hard coding problems |
| o4-mini | 200K tokens | Efficient reasoning, tool use, coding with lower cost |
OpenAI API FAQ
Everything you need to know about estimating and optimising your OpenAI API costs.
The OpenAI API gives developers access to OpenAI’s family of large language models, including GPT-4o, GPT-4.1, GPT-4o mini, and the reasoning-focused o-series (o3, o4-mini). It is used for chatbots, content generation, code assistance, data analysis, and agentic workflows.
The OpenAI API uses a token-based pricing model. You are billed for input tokens (your prompt and context) and output tokens (the model’s response) per million tokens. Different models have different per-token rates, with more capable models costing more. Reasoning models like o3 also bill for internal reasoning tokens as part of the output.
GPT-4o is OpenAI’s flagship multimodal model, excelling at general tasks, vision, and audio. GPT-4.1 is a newer model specifically optimised for coding, long-context instruction following, and agentic workflows, with a 1 million token context window. GPT-4o is better for multimodal tasks, while GPT-4.1 is better for software engineering and complex tool use.
o3 and o4-mini are OpenAI’s reasoning models. They use ‘chain of thought’ to solve complex problems in maths, science, and coding. o3 is the most capable but slower and more expensive, while o4-mini is faster and more cost-efficient. These models generate internal reasoning tokens that count toward output pricing.
Reasoning tokens are the internal ‘thinking’ tokens generated by o-series models (o3, o4-mini) as they work through complex problems step by step. These tokens are not shown in the final response but are billed as part of the output token count. This is why reasoning models can appear more expensive per query.
To optimise costs: 1) Use GPT-4o mini or GPT-4.1 nano for simple tasks. 2) Reserve o3 and GPT-4o for complex reasoning or multimodal needs. 3) Keep system prompts concise. 4) Use structured outputs to avoid unnecessary tokens. 5) Cache frequent responses. 6) For reasoning models, set reasoning_effort to ‘low’ or ‘medium’ when full reasoning isn’t required.
GPT-4.1 is OpenAI’s best model for coding, with top performance on SWE-bench and strong instruction following for agentic coding workflows. For extremely difficult algorithmic problems, o3 provides the strongest reasoning. For everyday coding assistance, GPT-4.1 mini offers excellent performance at a fraction of the cost.