AI Pulse · AI Trends Pulse

OpenAI just dropped prices on two of its GPT-5.6 models, and the cuts are steep enough that you should check what you’re paying for API calls. Luna, the smallest model in the lineup, fell 80 percent. It now costs $0.20 per million input tokens and $1.20 per million output tokens. Terra, the mid-tier option, dropped 20 percent to $2 and $12 per million tokens. The flagship Sol model held steady at $5 and $30, unchanged from the previous version.

The performance jump matters more than the price cut if you’re running anything that touches code. Sol Ultra, the top-end variant, hit 91.9 percent on OpenAI’s internal coding benchmark, up from 83.4 percent in GPT-5.5 according to the pricing announcement. That’s a meaningful gain if your workflows involve generating, reviewing, or debugging code at scale.

What this means for you depends on volume. If you’re burning through millions of tokens a month on automation, customer support, or document processing, Luna’s new rate makes it cheaper to handle high-volume, lower-complexity tasks without dropping quality. Terra slots in for work that needs more reasoning but doesn’t justify Sol’s cost. If you’ve been using Sol for everything, you can likely shift some workloads down and pocket the difference.

The catch is knowing which tasks belong where. Most companies don’t have clean visibility into how their AI spend breaks down by use case or model. That’s the kind of operational question we build into tools like the Omni Command Centre, where you can see cost per workflow and test whether a cheaper model still hits your accuracy bar. Pricing moves this fast, and unless you’re tracking it, you’re probably overpaying somewhere.