AI Models Now Compete on Price, Not Just Smarts
OpenAI wants to win by offering the cheapest way to get jobs done, as the AI race shifts from pure power to what actually matters at scale: cost, speed, and dependability.
@gdbTL;DR:
- What really matters now is how much each task actually costs to run.
- At big volumes, the price per job, how quick it is, and whether it works reliably beat any benchmark score.
- Developers are mixing models already, picking whichever is cheapest and fastest for the specific job.
- Open-source models and rivals are catching up fast, putting pressure on the big names to compete on price.
Headline
Brockman: OpenAI's goal is lowest price per task
Summary
Greg Brockman from OpenAI says they're going after the lowest cost for any given job. He's asking users for cases where competitors are cheaper or quicker on the same work. It's not just about benchmark numbers anymore—real cost per task is what matters when you run thousands of these.
Analysis
The game has changed. Once you're running at scale, it's all about the price per job, speed, and reliability. That puts pressure on the big labs because open models are getting closer in capability. OpenAI is okay with developers mixing and matching models based on what makes sense for each task, which is already how most people work anyway.