A quick reference for estimating how fast Codex usage drains your weekly credits based on model family, mode, reasoning effort, and your message pace.

codex-credit-burn-estimates.md

Codex weekly credit drain estimates (local usage)

Use these tables to estimate how many hours of active prompting it would take to drain a weekly allowance, assuming a steady pace.

How to scale these numbers to your setup

These tables assume:

• Weekly allowance = 1000 credits
• Pace = 20 messages/hour

To scale to your actual setup:

• If your weekly cap is W, multiply hours by (W / 1000).
• If your pace is P messages/hour, multiply hours by (20 / P).

---

A) GPT-5.2-Codex / GPT-5.1-Codex-Max / GPT-5.2

Assumption: local baseline ~5 credits/message

Mode	small	medium	high	very high
Chat	12.5 h	10.0 h	5.9 h	4.0 h
Agent	7.8 h	6.3 h	3.7 h	2.5 h
Agent (Full)	6.3 h	5.0 h	2.9 h	2.0 h

---

B) GPT-5.1-Codex-Mini

Assumption: local baseline ~1 credit/message

Mode	small	medium	high	very high
Chat	62.5 h	50.0 h	29.4 h	20.0 h
Agent	39.1 h	31.3 h	18.4 h	12.5 h
Agent (Full)	31.3 h	25.0 h	14.7 h	10.0 h

---

Plug-and-play formula

Let:

• W = your weekly credits cap (from Usage Dashboard)
• B = base credits/message (5 for big models, 1 for Mini)
• m = mode multiplier (1.0 / 1.6 / 2.0)
• r = reasoning multiplier (0.8 / 1.0 / 1.7 / 2.5)
• P = messages/hour

Then:

hours_to_drain ≈ W / (B * m * r * P)