# GPT-5.5 on the Cotera agent benchmark

> 5/5 at $1.21. Worth it when latency-vs-cost on the long-tail edge cases matters more than the median.

Source: https://cotera.co/benchmarks/models/azure-gpt-5.5

---

OpenAI's GPT-5.5 — the November 2026 thinking-heavy successor to GPT-5. The reasoning-tier OpenAI model most US enterprise agent stacks default to today.

**Score:** 5/5 benchmarks passed · **Cost across the matrix:** $1.21

## Strengths
- Five for five with the most variance-free run profile in the matrix. Two tool calls on Coding, two on Web Scraping, 3-7 on the others. Doesn't get distracted.
- Strongest performance on the Marketing rubric: ranked all five tools with mention counts and a one-line reason per item, no preamble.
- Per-cell cost is high but never wild. The most expensive single call was $0.435 (Marketing) — Sonnet's Marketing call was $1.82.

## Watch-outs
- It's a thinking model and it bills like one. ~3x the per-task cost of GPT-5 Mini for the same outcomes on Coding and Web Scraping.
- We didn't observe schema drift, but GPT-5.5's instinct is to reach for prose explanations when the task is ambiguous — the strict-JSON wrapper is doing real work here.
- Same short-horizon, no-adversarial caveats as the rest of the matrix.