Cheaper Frontier AI Changes the Math on What to Automate

The cost of running frontier AI just dropped again. OpenAI launched three GPT-5.6 models today — Sol, Terra, and Luna — and the most important number is not the benchmark score. It is the price. Terra matches GPT-5.5's performance at half the cost. That changes what is worth automating.

Here is what the pricing looks like. Sol, the flagship, costs $5 per million input tokens and $30 per million output tokens — the same price as GPT-5.5 was. Terra is $2.50 input and $15 output. Luna, the cheapest tier, is $1 input and $6 output. The gap between good enough and cheap enough has nearly closed. For any team running large volumes of AI calls, Luna and Terra are now worth a serious look.

Beyond price, GPT-5.6 adds a context window of 1.5 million tokens — about 43% bigger than before. That means longer documents, bigger codebases, and more complex multi-step tasks all fit in a single call. Sol also sets a new record on Terminal-Bench 2.1, a test that measures how well an AI handles real command-line work. Agentic reliability — how well the model completes long, multi-step jobs — has measurably improved.

The pattern is familiar: every few months, the same capability costs less. Teams that found AI automation too expensive six months ago may find it viable today. High-volume use cases — customer support, document processing, classification at scale — are now cheap to run with Luna. The practical question is no longer whether you can afford AI, but what you should automate next.