AI Pulse · Frontier Labs Watch

OpenAI just cut prices on GPT-5.6 by up to 75%, making it cheaper than GPT-4o mini in some cases. This isn’t a goodwill gesture. It’s a direct shot at anyone trying to run mid-tier agentic workloads on competing models, and it forces Anthropic and Google to respond on the cheap SKUs where most production agent volume actually happens, according to the original announcement.

The timing matters. Open-weight models like Kimi K3 and GLM-5.2 are trending right now, offering decent performance without the per-token bill. OpenAI’s move makes it harder for those alternatives to claim a cost advantage at scale. If you’re running agents that make thousands of API calls a day, a 75% price drop changes the math fast. It also means that if you built your stack around a competitor’s mid-tier model because it was cheaper last quarter, you now have to re-evaluate whether that’s still true.

For anyone running a business that depends on AI agents, this is a reminder that pricing is not stable. The floor keeps dropping, and the models keep shifting. What makes sense today might not next month. That’s why companies building serious AI operations are looking at systems that can swap models in and out without rewriting the whole pipeline. We build that kind of flexibility into the Omni Command Centre, so you’re not locked into one vendor’s pricing whims.

The broader point is this: the frontier labs are fighting hard for production workloads, not just benchmarks. If you’re still treating model selection as a one-time decision, you’re going to get caught out. The companies winning here are the ones treating it like infrastructure, where you can move fast when the landscape shifts.