OpenAI's New Model Leads With Price, Not Power

Eleven companies show up in OpenAI’s own launch materials today, and nearly every one of them says a version of the same thing: fewer tokens, lower cost, results that hold up or improve. That’s a different pitch than past model launches, which mostly led with raw capability.

OpenAI released GPT-5.6 on July 9, 2026, its first model family framed primarily around cost efficiency rather than intelligence alone. It ships in three tiers: Sol, the flagship; Terra, a mid-tier model competitive with the prior generation’s best; and Luna, the cheapest and fastest of the three.

Three tiers, one pitch

Pricing is public and specific: Sol runs $5 per million input tokens and $30 per million output tokens. Terra runs $2.50 and $15. Luna runs $1 and $6. OpenAI is explicit that the smaller tiers aren’t scaled-down afterthoughts; Luna reportedly outperforms OpenAI’s own previous flagship at a fraction of its cost.

The company also introduced two new settings for harder tasks. Max gives the model more time to reason before answering. Ultra runs several agents on the same problem in parallel, trading higher token spend for faster, stronger results on complex work.

Customers already building on it

The launch includes on-record quotes from Notion, Ramp, Shopify, Cisco, Canva, and Microsoft, along with two companies whose work maps directly onto agency back-office functions: Clio, a legal practice management platform, reported 14% fewer tokens with improved accuracy on legal research and document review. Basis, an accounting automation company, said the model showed meaningful gains in reasoning and autonomy on complex accounting work.

OpenAI also flagged stronger cybersecurity performance, including a jump on tests that measure turning a known software vulnerability into a working exploit, paired with expanded safeguards and a trusted-access program for verified security researchers.

Rollout starts now

GPT-5.6 is live today across ChatGPT, Codex, and the OpenAI API, with availability expanding gradually over the next 24 hours. Existing subscribers get access based on their current plan tier: paid ChatGPT users reach Sol through higher effort settings, while free and entry-level users default to Terra. Developers can call all three tiers directly through the API.

The model competes directly against Anthropic’s Claude and Google’s Gemini lines on the benchmark charts OpenAI published alongside the release, and lands the same week both competitors have been actively shipping their own updates, a sign the pace of releases across the major labs isn’t slowing down.