# GPT-5 Mini on the Cotera agent benchmark

> GPT-5 Mini quietly did the matrix in $0.38 — bring receipts the next time someone says "mini is too dumb."

Source: https://cotera.co/benchmarks/models/azure-gpt-5-mini

---

OpenAI's mini-tier in the GPT-5 family. Pitched as the cost-sensible option when full GPT-5 is overkill; in our matrix it earned that pitch.

**Score:** 5/5 benchmarks passed · **Cost across the matrix:** $0.376

## Strengths
- 5/5 pass at $0.38 total — the second-cheapest perfect score in the matrix.
- Aggressive but productive tool use on the Crunchbase brief: 10 tool calls, every one chained to the next, no thrashing.
- Coding and Web Scraping rubrics passed under $0.04 each — the cheap-and-correct quadrant most production agents should live in.

## Watch-outs
- Don't confuse this with full GPT-5. The mini variant trades raw reasoning for cost; tasks that need long planning chains will hit a wall before they hit budget.
- We didn't observe failures, but it's also the model in our suite with the least public documentation on context-window behavior over multi-hour chats.
- All the same caveats as the rest of the matrix: short-horizon tasks, no adversarial inputs, no multi-tenant stress.