Model Families and Selection

2 min read
Best ForContextInput/1MOutput/1M
GPT-5.2Coding, agentic tasks400K$1.75$14.00
GPT-5 / 5.1General purpose flagship400K$1.25$10.00
GPT-5 MiniBalanced quality and cost400K$0.25$2.00
GPT-5 NanoClassification, routing400K$0.05$0.40
GPT-4.1Long context (1M tokens)1M$2.00$8.00
GPT-4.1 MiniCost effective, long context1M$0.40$1.60
o3Complex reasoning, STEM200K$2.00$8.00
o4-miniFast reasoning200K$1.10$4.40
GPT-4oMultimodal with vision128K$2.50$10.00
GPT-4o MiniCheap multimodal128K$0.15$0.60

Reasoning token costs are hidden

o series models (o3, o4-mini, o1) generate internal reasoning tokens that are billed as output tokens but are not visible in the API response. A response showing 500 output tokens may actually consume 2,000 or more total tokens. Monitor o series usage carefully and avoid using these models for simple tasks where GPT-5 or GPT-4.1 would suffice.
1

Simple tasks

GPT-5 Nano or GPT-4o Mini ($0.05 to $0.15/1M input)

2

General purpose

GPT-5 Mini or GPT-4.1 Mini ($0.25 to $0.40/1M input)

3

Complex tasks

GPT-5.2, GPT-5, or GPT-4.1 ($1.25 to $2.00/1M input)

4

Deep reasoning

o3 or o4-mini ($1.10 to $2.00/1M input + reasoning tokens)