Provider guide
OpenAI API pricing and model catalog: use-case guide
OpenAI offers the broadest model catalog in the LLM API space: chat, reasoning, embedding, image generation, audio, and real-time models. This guide maps OpenAI's current pricing to common workloads so you can pick the right model tier without overpaying. For a cross-provider comparison, see the provider pricing comparison.
Pricing data sourced from our catalog. Check data sources for provenance and freshness.
Model tiers
OpenAI model tiers at a glance
Prices shown are per 1M tokens from the current catalog. Verify with OpenAI before making purchasing decisions.
| Tier | Examples | Input (per 1M) | Output (per 1M) | Best for |
|---|---|---|---|---|
| Budget | gpt-5-nano, gpt-5-nano-2025-08-07, gpt-4.1-nano | $0.0500 / 1M tokens | $0.4000 / 1M tokens | High-volume chat, classification, routing |
| Mid-range | gpt-3.5-turbo-1106, o3-mini, o3-mini-2025-01-31 | — | — | Balanced quality and cost |
| Premium | chatgpt-4o-latest, gpt-4o-2024-05-13, gpt-5.6 | — | — | Complex reasoning, long context |
| Reasoning | o1, o1-2024-12-17, ft:gpt-4-0613 | — | — | Multi-step tasks, math, code generation |
Cost example
What does a chat workload cost on OpenAI?
This is a rough estimate using the cheapest chat model. Actual costs depend on model choice, cache behavior, batch discounts, and rate tier. Use the calculator for your exact workload.
Use cases
Which OpenAI model for which workload?
Chat and customer support
For high-volume chat, use budget models like GPT-4o Mini. For complex multi-turn conversations, GPT-4o or GPT-5.4 provide better reasoning at higher cost.
RAG and document search
Use text-embedding models for vector search. Combine with chat models for answer generation. The embedding cost is often negligible compared to generation.
Summarization
For long documents, input tokens dominate the cost. Use budget models with large context windows to minimize per-document cost.
Image generation
DALL-E models are priced per image, not per token. Use the cost-per-image metric for budgeting, not token-based estimates.
Cost optimization
How to reduce OpenAI API costs
Use cache pricing. OpenAI offers cached input pricing at a fraction of the base rate. If your prompts have repeated prefixes (system prompts, RAG context), cached input can cut costs significantly. See the cache and batch pricing guide.
Use batch API. For non-interactive workloads (summarization, classification, data processing), the batch API offers lower prices with asynchronous processing.
Right-size the model. Don't use GPT-5.4 for classification tasks. Budget models handle straightforward tasks at a fraction of the cost. Use the model routing cascade framework.
Monitor output length. Output tokens cost 3-4x more than input tokens. Tighten system prompts and limit max_tokens where possible. See the output vs input pricing multiplier.
Full catalog
All OpenAI models in the database
| Model | Mode | Input (per 1M) | Output (per 1M) | Context |
|---|---|---|---|---|
| text-embedding-3-small | embedding | $0.0200 / 1M tokens | — | 8k |
| text-embedding-ada-002 | embedding | $0.1000 / 1M tokens | — | 8k |
| text-embedding-ada-002-v2 | embedding | $0.1000 / 1M tokens | — | 8k |
| text-embedding-3-large | embedding | $0.1300 / 1M tokens | — | 8k |
| gpt-5-nano | chat | $0.0500 / 1M tokens | $0.4000 / 1M tokens | 128k |
| gpt-5-nano-2025-08-07 | chat | $0.0500 / 1M tokens | $0.4000 / 1M tokens | 128k |
| gpt-4.1-nano | chat | $0.1000 / 1M tokens | $0.4000 / 1M tokens | 33k |
| gpt-4.1-nano-2025-04-14 | chat | $0.1000 / 1M tokens | $0.4000 / 1M tokens | 33k |
| gpt-4o-mini | chat | $0.1500 / 1M tokens | $0.6000 / 1M tokens | 16k |
| gpt-4o-mini-2024-07-18 | chat | $0.1500 / 1M tokens | $0.6000 / 1M tokens | 16k |
| gpt-4o-mini-audio-preview | chat | $0.1500 / 1M tokens | $0.6000 / 1M tokens | 16k |
| gpt-4o-mini-audio-preview-2024-12-17 | chat | $0.1500 / 1M tokens | $0.6000 / 1M tokens | 16k |
| gpt-4o-mini-search-preview | chat | $0.1500 / 1M tokens | $0.6000 / 1M tokens | 16k |
| gpt-4o-mini-search-preview-2025-03-11 | chat | $0.1500 / 1M tokens | $0.6000 / 1M tokens | 16k |
| ft:gpt-4.1-nano-2025-04-14 | chat | $0.2000 / 1M tokens | $0.8000 / 1M tokens | 33k |
| gpt-5.6-luna | chat | $0.2000 / 1M tokens | $1.2000 / 1M tokens | 128k |
| gpt-5.4-nano | chat | $0.2000 / 1M tokens | $1.2500 / 1M tokens | 128k |
| gpt-5.4-nano-2026-03-17 | chat | $0.2000 / 1M tokens | $1.2500 / 1M tokens | 128k |
| ft:gpt-4o-mini-2024-07-18 | chat | $0.3000 / 1M tokens | $1.2000 / 1M tokens | 16k |
| gpt-3.5-turbo | chat | $0.5000 / 1M tokens | $1.5000 / 1M tokens | 4k |
Showing 20 of 145 models. View all OpenAI models.
Related guides
Continue reading
Cross-provider pricing comparison
How OpenAI pricing compares against Anthropic, Google, Mistral, and DeepSeek.
Hidden costs of LLM APIs
Rate limits, latency, evaluation overhead, and vendor risk beyond per-token pricing.
Model routing cascade
When to use budget vs premium models across providers.
Cache and batch pricing guide
How cached input and batch pricing change OpenAI cost estimates.
Tools
Estimate your OpenAI costs
Disclaimer
Prices are sourced from public provider documentation and the site's model catalog. Verify with OpenAI before making purchasing decisions. This article is not affiliated with OpenAI.