Calculate generation limits and costs dynamically across OpenAI's API. Supporting GPT-5.4, o3 reasoning models, and standard GPT-4 architectures.
Prices update in real-time. Results under $0.0001 are rounded up.
Category: Flagship
Input: $2.5 | Cached: $0.25 | Output: $15
Category: Flagship
Input: $0.75 | Cached: $0.075 | Output: $4.5
Category: Flagship
Input: $0.2 | Cached: $0.02 | Output: $1.25
Category: Flagship
Input: $30 | Cached: N/A | Output: $180
Category: Flagship
Input: $1.75 | Cached: $0.175 | Output: $14
Category: Flagship
Input: $1.25 | Cached: $0.125 | Output: $10
Category: Flagship
Input: $1.25 | Cached: $0.125 | Output: $10
Category: Flagship
Input: $0.25 | Cached: $0.025 | Output: $2
Category: Reasoning
Input: $2 | Cached: $0.5 | Output: $8
Category: Reasoning
Input: $20 | Cached: N/A | Output: $80
Category: Reasoning
Input: $1.1 | Cached: $0.55 | Output: $4.4
Category: Reasoning
Input: $1.1 | Cached: $0.275 | Output: $4.4
Category: Reasoning
Input: $15 | Cached: $7.5 | Output: $60
Category: Reasoning
Input: $150 | Cached: N/A | Output: $600
Category: Reasoning
Input: $1.1 | Cached: $0.55 | Output: $4.4
Category: Previous Gen
Input: $2.5 | Cached: $1.25 | Output: $10
Category: Previous Gen
Input: $0.15 | Cached: $0.075 | Output: $0.6
Category: Previous Gen
Input: $10 | Cached: N/A | Output: $30
Category: Previous Gen
Input: $0.5 | Cached: N/A | Output: $1.5
Category: Specialized
Input: $10 | Cached: $2.5 | Output: $40
Category: Specialized
Input: $2 | Cached: $0.5 | Output: $8
Category: Specialized
Input: $3 | Cached: N/A | Output: $12
Regional processing endpoints (data residency requirements) typically incur a 10% price uplift for Flagship generation models like GPT-5.4, GPT-5.4 Mini, and GPT-5.4 Pro.
Models like o1, o3, and their mini variants consume invisible "reasoning tokens" during execution. These are billed at standard output rates, meaning your generated response cost includes both the visible content and the underlying thought process steps.
Connecting generation tasks to web search incurs an additional flat rate per 1k iterations plus standard execution costs. For Reasoning models, search preview costs $10.00 / 1k calls. Non-reasoning models run at $25.00 / 1k calls.
For capable models like GPT-4o and GPT-5.4, OpenAI can reuse cached blocks of tokens across multiple identical prompt fragments, applying massive discounts to the input cost. For instance, GPT-5.4 features a 90% discount (input drops from $2.50 to $0.25) when matching cached data.
The tokens consumed to understand, generate, and process tool calls are tracked precisely alongside input and output parameters, mapped at the selected model's per-token rates. However, specialized tools like File Search incur discrete "Tool call" access costs (e.g., $2.50 / 1k calls).
Unlike standard text, multimodal features such as sending images or video frames are mapped individually using grid constraints. A standard High-Res image tile is roughly 170 tokens each plus an 85 token base footprint constraint per image.
OpenAI offers a specialized computer-use-preview interface priced affordably at $3 / 1M input tokens, supporting automated web actions and software-layer integrations autonomously guided by visual context grids.