OpenAI slashed prices with the release of its new GPT-4.1 model family, offering significant savings compared to previous versions. The company’s latest AI models come in three flavors: standard GPT-4.1 (the smartest one), mini (balanced), and nano (fastest and cheapest). It’s a pretty obvious play to get more businesses on board. Who doesn’t love saving money?

OpenAI’s new GPT-4.1 lineup slashes prices while offering smarter AI. Three tiers means something for everyone’s budget.

The pricing structure is straightforward. GPT-4.1 costs $2 per million tokens for input and $8 per million for output. That’s way cheaper than the original GPT-4, which was practically bleeding wallets dry at $30 input/$60 output for the same token count.

Mini version? Just $0.40 input and $1.60 output. Nano is even cheaper at $0.10 and $0.40 respectively. Talk about a bargain.

For companies watching their budgets—and who isn’t these days?—OpenAI introduced a Batch API. It processes requests over 24 hours instead of instantly, slashing costs by 50%. The Batch API allows for sending asynchronous groups of requests, ideal for use cases not requiring immediate responses.

Yeah, you’ll wait longer, but your accountant will thank you. A million output tokens on GPT-4.1? Regular price: $8. Batch price: $4. Do the math.

Enterprise users get even more goodies. Custom pricing, beefed-up security with SOC 2 compliance and encryption, admin controls with SSO. Plus dedicated support when things inevitably go sideways.

Azure OpenAI Service offers similar enterprise perks with throughput options for the impatient types.

Smart users are already finding ways to stretch their token budgets. Shorter prompts. Less fluff. No “please” and “thank you” for the AI—it doesn’t have feelings anyway.

Removing whitespace from JSON, setting token limits, tuning temperature to zero for more predictable outputs.

The token-based pricing model remains unchanged. You pay for what you use, roughly 4 characters per token. Output costs more than input, because apparently thinking is expensive. The model features a 128K context window, allowing it to process approximately 300 pages of text at once.

Cached inputs offer additional savings for repeat requests. Bottom line? OpenAI wants your business, and they’re willing to charge less to get it.