TechCrunch: OpenAI releases GPT-5.6 family with tiered pricing and ChatGPT Work

On July 9, 2026, OpenAI released its GPT-5.6 model family in three tiers: Sol (high-performance), Terra (mid-range), and Luna (budget). Pricing is published per million tokens — Sol at $5 input / $30 output, Terra at $2.50 / $15, Luna at $1 / $6. OpenAI claims Sol delivers 54% more token efficiency on coding tasks compared to previous generations, and outperforms Anthropic’s Fable 5 by 2.8 points on coding benchmarks while using roughly half the tokens at one-third the cost.

Alongside the model release, OpenAI launched ChatGPT Work — a desktop, web, and mobile companion designed for enterprise teams to handle document drafting, presentations, and routine coordination. This positions OpenAI more directly in the enterprise productivity space that Anthropic has built significant revenue in over the past year.

For product managers who build on AI APIs, the tiered structure has immediate practical relevance. Routing high-stakes reasoning tasks to Sol, standard workflows to Terra, and high-volume low-complexity operations to Luna can meaningfully reduce per-request costs without changing the user experience. Token-efficiency benchmarks also signal a broader shift: the AI industry is moving toward evaluating models not only by performance scores but by output per dollar, which is the metric that matters most once AI features move from prototypes to production at scale.

The ChatGPT Work launch is worth watching for a different reason. OpenAI is moving from API provider to workplace product — the same surface area that Anthropic, Google, and Microsoft have been competing on. For teams that make infrastructure decisions about which AI vendors to consolidate around, the competitive picture in this segment will look different by the end of 2026.