Claude Opus 5.5 now available in Model ML
5 MIN. READ
Claude Opus 5.5 is live in Model ML.
Our evals team ran it pre-release through Model ML's Composite, the leading benchmark for AI in financial services, against both closed-source and open-source models.
A few things stood out:
Opus 5.5 is a clear improvement over Opus 5, scoring higher across all five categories. Against the wider field, it consistently ranks third or fourth. GPT-6 Astra still leads on raw quality in four of the five categories, with its widest lead over Opus 5.5 coming on multi-document intelligence, where it scores about 21% higher in accuracy.
While cheaper than Opus 5 on four out of five categories, Opus 5.5 still remains on the premium end of the frontier. The starkest example is financial workflows, where Gemini 3.8 Flash scored 53% higher while costing roughly 48% less than Opus 5.5.
The gap between open-source and closed-source models is closing. For example, DeepSeek 4.1 Flash outscores Opus 5.5 on financial workflows by about 20% at about 7% of the cost. On single document intelligence, it’s about the same level of performance Opus 5.5 at ~12% of the cost.
No single model wins on price and performance at once, which is the whole reason we stay model-agnostic: Model ML routes each task to whatever's best for it.
Try Claude Opus 5.5 in Model ML today.
