OpenAI

Lineup

Default to GPT-5.1 ($1.25/$10 per M in/out, 400K context) for general work — it's the best price-to-capability ratio in the current lineup. Need coding-agent behavior specifically? GPT-5.1-Codex is priced the same. For high-volume, low-stakes tasks, GPT-5.4 Nano ($0.20/$1.25) or the older GPT-4.1 Nano ($0.10/$0.40) cut cost further, though gpt-oss-20b ($0.029/$0.14) and GPT-5 Nano ($0.05/$0.40) are cheaper still if you don't need OpenAI's newest weights. For zero-cost testing, gpt-oss-120b:free and gpt-oss-20b:free run at no charge. At the top end, o1-pro ($150/$600) and GPT-5.5 Pro ($30/$180) are reasoning/heavy-compute tiers reserved for problems where cost isn't the constraint.

Questions

What's the price range for OpenAI models on OpenKey?

Input pricing spans from free (gpt-oss-120b:free, gpt-oss-20b:free) to $150/M on o1-pro, the most expensive model in the lineup ($150 input × 1.03 = $154.50 on OpenKey). Output tops out at $600/M on o1-pro. Most day-to-day models — GPT-5.1, GPT-5, GPT-4.1 — sit in the $0.25–$2.50 input range, with OpenKey adding a flat 3% on top of OpenAI's list price.

Which OpenAI model is cheapest, and which is best for general use?

gpt-oss-20b is the cheapest paid option at $0.029/M input and $0.14/M output ($0.0299/$0.1442 on OpenKey after the 3% fee); gpt-oss-120b:free and gpt-oss-20b:free cost nothing. For general-purpose work, GPT-5.1 ($1.25/$10 per M, 400K context) is the best-balanced flagship — GPT-5.5 Pro and o1-pro cost far more and are meant for heavier reasoning loads, not everyday use.

Do I need a separate OpenAI account to use these models?

No — OpenKey gives you one API key across all 64 OpenAI models plus every other lab on the platform (329 models across 52 labs total). You're billed at OpenAI's list price plus a flat 3% fee, with no separate OpenAI account or billing relationship required.