Framer Launches GPT 5.6 Sol, Terra, and Luna Models with Improved Speed and Cost Options.

Framer has rolled out GPT 5.6 in three distinct configurations: Sol, Terra, and Luna. Each targets a different use case within the platform's AI-powered design workflow.

Sol is positioned as the strongest creative option. According to Framer's internal benchmarks, it scored 100% on the company's hardest eval set, a first for any model tested. It costs the same in credits as GPT 5.5, making it a straight upgrade for users who prioritize output quality over speed or volume.

Terra delivers GPT 5.5-level results at 0.6× the credit cost. Framer suggests it for large-scale work like site audits, redesigns, and consistency passes where quality matters but budget constraints are real.

Luna runs twice as fast as any previous model in Framer and uses only 0.4× the credits of GPT 5.5. It's built for high-volume tasks: CMS updates, cross-site find-and-replace, and bulk operations where speed and cost matter more than creative nuance.

The update follows Framer's recent launch of Fable 5, which scored 83% overall in the company's evals and led on design at 81%. Fable sits at 2× the credit usage of GPT 5.5, roughly 3.3× more expensive than Sonnet 5.

Framer's benchmark suite tests models on layout accuracy, visual coherence, and adherence to prompts. Sol's 100% score on the hardest internal eval suggests it handles complex creative briefs with minimal revision cycles. That matters in practice: fewer back-and-forth edits mean faster iteration, even if the model itself isn't the fastest runner.

Terra's positioning is straightforward: same intelligence as GPT 5.5, lower cost. For teams running dozens of design tasks per week, the 40% credit savings compound quickly. Framer's messaging suggests Terra as the default for production work where quality can't slip but budget is finite.

Luna's 2× speed advantage comes with trade-offs. Framer explicitly recommends it for structured, repetitive tasks rather than open-ended design. The 0.4× credit multiplier makes it the cheapest option in the lineup by a wide margin, though Sonnet 5 still undercuts it on cost per task in some scenarios.

Framer has been steadily expanding its model roster. Sonnet 5 launched two weeks ago with stronger design instincts than Sonnet 4.6 and lower costs. Fable 5 arrived around the same time, scoring highest on design-specific tasks. GPT 5.6's three-model split is the clearest example yet of Framer treating AI selection as a workflow decision, not a one-size-fits-all toggle.

The company has also been adding workflow guardrails, like pre-launch reviews to catch errors before publishing. That suggests Framer is thinking about AI as part of a production pipeline, not just a creative assistant.

All three GPT 5.6 models are available now in the agent model picker. Users can switch between them mid-project depending on the task at hand. Framer has published full benchmark results for transparency on performance and cost trade-offs.