GPT-5.6 Sol, Terra and Luna now available in Model ML

Model ML, along with other leading agent harness companies including Lovable and Notion, partnered with OpenAI to test GPT-5.6 before its public release.

Model ML ran GPT-5.6 inside its agent harness against The Model ML Composite, its benchmark for financial services. The evaluation spanned 20 of Model ML's most challenging client workflows, thousands of generated outputs, and 32 composite criteria.

Compared to Claude Fable 5, GPT-5.6 Sol used 39% fewer tokens to create PowerPoint presentations, and the results were more polished and legible, with clearer, more accurate data visualisations. Model ML also tested the outputs with human finance professionals, who found the decks required far less rework before they were ready to share.

"GPT-5.6 Sol is the first model we've evaluated that consistently generates decks ready for real work. Across 20 challenging client workflows and hundreds of decks, it used 39% fewer tokens per deck than Fable 5 while producing more polished, legible decks with clearer, more accurate data visualizations that required less rework before sharing."

— Chaz Englander, Co-Founder & CEO at Model ML

The full findings can be accessed in OpenAI’s GPT-5.6 launch announcement.