OpenAI Drops GPT-6 Sol and Luna with 50% Lower API Prices

Something significant happened in enterprise AI on September 22, 2026. Not one major AI provider cut its prices — two did, on the same day.

OpenAI released GPT-6 Sol and GPT-6 Luna, two new models that position themselves as the workhorse tier beneath the flagship GPT-6 Astra. Hours earlier, Anthropic had released Claude Opus 5.5 at 40% lower cost than its predecessor. The timing was not accidental. The AI price war is real, it is accelerating, and it is changing the economics of building with AI.

What OpenAI Just Released

GPT-6 Sol and GPT-6 Luna are mid-tier and high-volume models designed for the work that enterprises actually do most — not the occasional frontier reasoning task, but the daily grind of coding, data analysis, summarization, and extraction.

GPT-6 Sol is priced at $2 per million input tokens and $10 per million output tokens. That is exactly half what GPT-5.6 Sol cost ($4 and $20 per MTok). OpenAI positions Sol for recurring developer and knowledge work: building features, reviewing code, debugging, analyzing data. The model makes roughly half as many errors as its predecessor.

GPT-6 Luna comes in at $0.10 and $0.50 per million tokens — a rounding error compared to what enterprise AI used to cost. Luna is built for high-volume, well-defined jobs: summarization, extraction, answering structured questions. At higher effort settings, it reportedly matches GPT-5.6 Sol at about a hundredth the cost.

Both models share a 1.05 million token context window, with 922K input capacity and 128K max output. They are available across ChatGPT Work, Codex, and the API for Plus, Pro, Business, Enterprise, and Edu tiers.

The Broader Context: A 40% Cut Hit on Both Sides

Anthropic’s Claude Opus 5.5 — also released September 22 — brought a 40% reduction in typical workload costs compared to Opus 5. At $4/$20 per MTok with a 30% speed improvement, it performs at the level of Claude Fable 5.1 on most tasks.

The result: in one 24-hour window, the two dominant enterprise AI providers both dropped the price of running serious AI workloads by 40-50%.

That is not a coincidence. It reflects a structural shift in how the frontier AI market is developing. The underlying compute costs have fallen, inference efficiency has improved, and competition between providers has intensified. What cost $100 per million tokens a year ago costs a fraction of that now.

What This Means for Business

For business owners and operations leaders still building their AI case internally, this changes the numbers.

The classic objection to AI agent deployments has been cost. A process that runs thousands of AI calls per day could rack up serious API bills at 2024 or 2025 pricing. That objection is getting weaker by the month.

At Luna’s pricing ($0.10/$0.50 per MTok), routine extraction and summarization tasks become trivially cheap to automate. At Sol’s pricing ($2/$10), more complex agentic workflows — the kind that handle multi-step reasoning across documents, systems, and decisions — become financially viable at scale.

Three practical shifts businesses should consider:

Re-tier your AI workloads. Most AI applications do not need the most powerful (and most expensive) model for every call. Luna handles structured extraction; Sol handles complex reasoning; Astra (or Fable 5.1) handles the genuinely hard problems. Running the right model for each task is now one of the main cost levers in AI operations.

Revisit your build decision. If a custom AI workflow looked too expensive six months ago, the math may have changed. What once cost $50,000/year in AI compute might now cost $20,000. Run the numbers again before assuming it’s not viable.

Lock in good architectural habits now. The models are getting cheaper and more capable simultaneously. The teams that will win are not the ones waiting for the “right time” — they are the ones building the foundation now, so they can swap in better models as they arrive without rebuilding from scratch.

What EDNA Sees in This Trend

The democratisation of frontier AI capability is accelerating. Cost is no longer the barrier it once was. The new barrier is implementation: having the right team, the right architecture, and the right strategy to actually deploy AI agents that run reliably in production.

That is precisely where Enterprise DNA works with businesses. Whether you are evaluating AI platforms, building your first internal workflow agent, or scaling a deployment that’s already working, the question is less “can we afford AI” and more “do we have the capability to use it well.”

The price war is good news for businesses. But price alone does not create transformation.

---

Enterprise DNA helps organisations deploy AI agents, build custom AI tools, and develop internal data capability. If you want to understand what the new AI pricing landscape means for your specific situation, book a discovery call.