aiexpert
Home / News / Brief
Research · Aug 10, 2026, 02:03 PM · 3 sources

GPT-5.6 Sol powers Model ML finance agent; 39% fewer tokens per investment deck than Fable 5

Model ML's finance agents powered by GPT-5.6 Sol produced investment decks using 39% fewer tokens per deck than Fable 5, while generating more polished, legible decks with clearer, more accurate data visualizations that required less rework before sharing.

A bespoke tearsheet that once took an analyst about an hour to assemble now takes about five minutes using GPT-5.6 Sol, which improved knowledge work outputs across domains and excels at equations, financial models, and typography, spacing, hierarchy and worksheet layout. Model ML's end-to-end workflow carries a finance request through research, analysis, and finished PowerPoint or Excel output, with agent routing each step to the model best suited for it.

In independent FinanceReasoning benchmarks, GPT-5.6 Sol matches Claude Fable-5's 90.34% accuracy at $3.85 per run against Fable-5's $10.05 (38% of Fable-5's cost), while GPT-5.6 Sol-Pro records the benchmark's highest accuracy at 90.76% at $16.35 per run. For finance builders: Sol's token efficiency and turnaround-time compression signal meaningful cost savings in document-heavy workflows. The accuracy gap versus Pro becomes irrelevant when the base tier hits your business thresholds on correctness.

Sources

Everything this brief rests on
  1. 01 Primary source openai.com
  2. 02 aimultiple.com aimultiple.com “gpt-5.6-sol matches claude-fable-5's 90.34% at $3.85 per run against fable-5's $10.05, at 38% of fable-5's cost”
  3. 03 artificialanalysis.ai artificialanalysis.ai “GPT-5.6 Sol (max) offers a similar level of intelligence to Claude Fable 5 at approximately one third of the cost”