Model ML Turbocharges Finance Work with GPT-5.6 Sol

Startup Model ML has announced a significant leap in financial workflow automation, powered by OpenAI's latest model, GPT-5.6 Sol. The company, which builds AI agents to handle the complex "last mile" of finance work, is seeing dramatic efficiency gains and improved output quality. This development, detailed on OpenAI News, points to a future where AI assistants tackle intricate tasks with near-human polish.

For finance professionals, the "last mile" often involves laborious tasks like reconciling data, formatting complex spreadsheets, linking every claim to its source, and ensuring final documents are editable and ready for executive review. Model ML, founded by brothers Arnie and Chaz Englander, emerged from their own experience investing and building software to streamline these processes. Their platform uses a core agent that orchestrates various tools, including GPT-5.6 Sol, to manage workflows from initial request through research, analysis, and final deliverable.

Sharper Outputs, Fewer Tokens

The integration of GPT-5.6 Sol has yielded impressive results. In Model ML's internal benchmark, "Composite," GPT-5.6 Sol demonstrated a 36% reduction in tokens per Excel workbook compared to Opus 5, and a 21% reduction per PowerPoint deck compared to Fable 5. Beyond raw token counts, the model's ability to produce "professional-ready" outputs has dramatically improved. For PowerPoint decks, GPT-5.6 Sol achieved a 43.3% professional-readiness rate, a 16.6 percentage-point lead over Opus 5. This means more generated decks are immediately suitable for substantive review, a critical bottleneck in finance.

"Earlier models could do the work of an analyst, but the user would have to clearly break down the task, specifically what it wanted the output to look like," Chaz Englander, Co-founder and CEO at Model ML, stated. "With GPT-5.6 Sol, we're finding that the agent gets far closer to the final output." This enhanced capability allows finance teams to offload the drudgery of formatting and data linking, freeing them to concentrate on higher-value activities like refining assumptions and sharpening strategic messaging.

Solving the Last Mile Challenge

Model ML's platform is designed to be "surface-agnostic," meaning users can initiate tasks in email, the Model ML app, or Microsoft Office plug-ins, maintaining continuity. For investment decks, it can transform a brief and source material into an editable PowerPoint presentation. For Excel tasks, it can start with client templates, gather data, build complex formulas across multiple sheets, and apply finance-specific formatting. This end-to-end automation drastically cuts down preparation time. At one global asset manager, a bespoke tearsheet that previously took an analyst an hour now takes about five minutes.

The company's evaluation framework, Composite, rigorously tests AI models across finance workflows. It assesses not just the accuracy of numbers but also the traceability of sources, formula integrity, structural correctness, and visual quality. GPT-5.6 Sol's performance in these benchmarks, particularly its ability to produce editable files with linked sources and recalculating workbooks, sets a new standard for AI in financial services. StartupHub.ai data shows Model ML with a score of 35/100, placing it in a competitive space against giants like Databricks (82/100) and Hugging Face (76/100), though its focus on specialized finance workflows offers a distinct niche. Model ML has verified financials, having raised $75M in a Series B round in 2025.

Building for the Future of Finance

Model ML's customers are increasingly demanding outputs that remain connected to their underlying models and source material, enabling interactive reports that can be updated or locked at specific moments. A reviewer might click on a figure in an investment summary, trace it back to the financial model, and even interact with the AI agent from that same page. Englander notes that traditional office software was designed for manual creation, but AI is fundamentally changing this assumption, pushing software itself towards more dynamic, integrated experiences.

The partnership between Model ML and OpenAI, involving on-site sessions to refine agent planning and tool selection, highlights a collaborative approach to pushing AI capabilities. By providing agents with specialized toolkits for data integration, document editing, and code execution, Model ML ensures its AI stays focused and effective. This focus on practical, review-ready outputs for complex financial tasks underscores the maturation of AI agents beyond simple task execution into sophisticated co-pilots for knowledge work.