AI Pulse · Frontier Labs Watch

OpenAI has cut API prices for its GPT-5.6 Sol model again. Short-context input is dropping from $5 to $4 per million tokens, while output falls from $30 to $20 per million. The new rates are guaranteed through November 21. This follows July cuts to two related models, Luna by 80% and Terra by 20%, making three pricing moves across the same model family in four weeks.

For a business using AI in customer support, document processing, reporting, or internal assistants, output pricing is often where the bill grows fastest. A one-third reduction on output can materially change the economics of workflows that generate long responses, summaries, analysis, or drafted content. It may also make it practical to move work from a cheaper but less capable model into a stronger one, provided the quality gain is worth it.

The bigger signal is competition. These cuts are being framed as a response to Anthropic and lower-priced Chinese open models. That suggests genuine pressure on model margins, not a temporary promotional offer. Don’t assume today’s provider pricing will hold, though. Build your AI workflows so you can monitor usage, compare model quality and cost, and switch where it makes commercial sense. This is the kind of thing we build into an AI command centre, giving operators a clearer view of where AI spend is going and what it’s producing.