Cheapest GPT API in 2026

Every GPT model ranked by price. Find the cheapest option for your use case — from free-tier to flagship.

Updated July 5, 2026 · 10 OpenAI models compared

OpenAI now offers 10 GPT models at wildly different price points. Choosing the wrong one can cost you 50x more than necessary. Here's the complete ranking from cheapest to most expensive.

GPT Models Ranked by Price (Cheapest First)

#ModelInputOutputBest For
1GPT-oss 20B$0.08/M$0.35/MSimple tasks, classification
2GPT-4o mini$0.15/M$0.60/MChat, summaries, drafts
3GPT-5.4 nano$0.20/M$1.25/MFast inference, high volume
4GPT-oss 120B$0.15/M$0.60/MCode, reasoning (open-weight)
5GPT-5 mini$0.25/M$2.00/MBalanced quality/cost
6GPT-5.4 mini$0.75/M$4.50/MComplex tasks, good quality
7GPT-4o$2.50/M$10.00/MMultimodal, vision
8GPT-5$1.25/M$10.00/MHigh-quality general tasks
9GPT-5.4$2.50/M$15.00/MAdvanced reasoning
10GPT-5.5$5.00/M$30.00/MFlagship, most capable

💡 The Big Insight

GPT-oss 20B is 62x cheaper than GPT-5.5 for input tokens. If you're using GPT-5.5 for tasks that don't require flagship quality, you're overpaying by thousands of dollars per year.

Real Cost Examples

Here's what you'd pay per month at different usage levels (assuming 75% input, 25% output token ratio):

Model100K tokens/mo1M tokens/mo10M tokens/mo
GPT-oss 20B$0.01$0.10$1.03
GPT-4o mini$0.02$0.19$1.88
GPT-5 mini$0.06$0.63$6.25
GPT-4o$0.19$1.88$18.75
GPT-5.4$0.34$3.44$34.38
GPT-5.5$0.69$6.88$68.75

Which GPT Model Should You Use?

For chatbots and customer support: GPT-4o mini ($0.15/M input) handles most conversations well. Upgrade to GPT-5 mini only if you need better reasoning.

For code generation: GPT-5 mini ($0.25/M input) or GPT-5.4 mini ($0.75/M input) for complex code. GPT-oss 120B is a cheap open-weight alternative.

For document analysis: GPT-4o ($2.50/M input) if you need vision. GPT-5 ($1.25/M input) for text-only long documents.

For research and complex reasoning: GPT-5.4 ($2.50/M input) or GPT-5.5 ($5.00/M input) only when you need maximum capability.

Find YOUR cheapest model

Enter your current model and monthly spend. We'll show you exactly how much you could save — and which model to switch to.

Run Free Cost Audit →

How to Reduce Your GPT API Costs

1. Right-size your model. Most GPT-5.5 usage can be handled by GPT-5 mini at 95% less cost. Run an audit to find out.

2. Use caching. OpenAI's prompt caching reduces input costs by 50% for repeated prompts.

3. Batch your requests. Batch API gives 50% off for non-urgent workloads.

4. Monitor for new models. OpenAI launches cheaper models regularly. The GPT-5.4 nano ($0.20/M) didn't exist 6 months ago.

🔒 Pro Tip: Set Up Price Alerts

APIpulse Pro monitors all GPT models 24/7. When OpenAI drops prices or launches a cheaper model, you get an email with exact savings and migration code ready to paste. Get Pro for $19 →

Frequently Asked Questions

Which GPT model is the cheapest?
GPT-oss 20B is the cheapest at $0.08/M input and $0.35/M output. For mainstream models, GPT-4o mini is cheapest at $0.15/M input and $0.60/M output.
How much does GPT-5.5 cost per token?
GPT-5.5 costs $5.00/M input tokens and $30.00/M output tokens. GPT-5.5 Pro costs $30.00/M input and $180.00/M output.
Is GPT-4o mini cheaper than Claude Haiku?
Yes. GPT-4o mini ($0.15/M input, $0.60/M output) is about 6x cheaper than Claude Haiku 4.5 ($1.00/M input, $5.00/M output) for input tokens.
How much can I save by switching GPT models?
Switching from GPT-5.5 to GPT-5 mini saves about 95%. Switching from GPT-4o to GPT-4o mini saves about 90%. A developer spending $500/mo on GPT-5.5 could save $475/mo.
Should I always use the cheapest model?
No. Match the model to the task. Use GPT-oss 20B for simple classification, GPT-4o mini for chat, GPT-5 for complex reasoning, and GPT-5.5 only for tasks requiring maximum capability. Most users overpay by using flagship models for simple tasks.