Developer cost model
Codex Cost Calculator
Codex costs depend on local task size, cloud task duration, code review depth and the model selected for the workflow.
No login. No AI API calls. Calculations run locally in your browser.
Last updated: 2026-08-06. Estimate only. This is not a billing statement.
Inputs
AI coding agent cost calculator
Model subscription fees, BYOK API usage, cache reads, cache writes and team workload in one local calculator.
Results
No login. No AI API calls. Calculations run locally in your browser.
OpenAI / GPT-5 (legacy). Last updated: 2026-08-06
Estimated monthly cost
$46.01
Estimated daily cost
$2.09
Cost per task
$0.42
Estimated yearly cost
$552.10
Token volume
Monthly input tokens
9.90M
Monthly output tokens
1.32M
Monthly cache read tokens
3.46M
Monthly cache write tokens
792.0K
Cost risk level
Medium
Main cost driver: subscription fee
Suggested ways to reduce cost
- Keep task context bounded so input tokens do not grow silently.
- Use explicit task boundaries before starting an agent run.
- Prefer patch-oriented output over broad explanations.
- Review high-cost sessions and reuse successful prompts.
TokenPatch
Want to measure cost per applied patch instead of raw token usage? Try TokenPatch.
Workflow assumptions
TokenPatch
Want to measure cost per applied patch instead of raw token usage? Try TokenPatch.
Pricing data
Model price assumptions
Pricing data is stored locally in the site configuration and should be treated as a planning estimate. Verify official provider pages before budgeting.
| Provider | Model | Input / 1M | Output / 1M | Cache read / 1M | Cache write / 1M | Last updated | Source |
|---|---|---|---|---|---|---|---|
| Anthropic | Claude Sonnet 4.6 Claude Sonnet 5 available at $2/$10 introductory through Aug 31; moves to $3/$15 Sep 1. Claude Code plan limits and enterprise contracts can differ from API token pricing. | $3 | $15 | $0.3 | $3.75 | 2026-08-06 | Official |
| Anthropic | Claude Opus 4.8 Opus is usually the high-cost path for long-running agent sessions. Sonnet is cheaper for most coding workflows. | $5 | $25 | $0.5 | $6.25 | 2026-08-06 | Official |
| OpenAI | GPT-5 (legacy) Legacy pricing. OpenAI recommends GPT-5.4 ($2.50/$5 standard) or GPT-5.6 ($5/$10) for new workloads. Codex plans and API billing differ. | $1.25 | $10 | $0.125 | $0 | 2026-08-06 | Official |
| OpenAI | GPT-5.4 mini Standard pricing. Good for bounded coding tasks where cheaper iterations matter more than peak reasoning. | $0.75 | $4.5 | $0.075 | $0 | 2026-08-06 | Official |
| Gemini Pro Gemini Pro-class paid-tier pricing for prompts at or below 200k tokens. Verify current pricing. | $2 | $12 | $0.2 | $0 | 2026-08-06 | Official | |
| DeepSeek | DeepSeek V4 Pro V4 Pro is the default tokenpatch executor (BYOK). DeepSeek has announced plans to raise pricing; verify before budgeting. V4 Flash ($0.14/$0.28) is a cheaper alternative. | $0.435 | $0.87 | $0.003625 | $0 | 2026-08-06 | Official |
| DeepSeek | DeepSeek V4 Flash Cheaper alternative. DeepSeek has announced plans to raise overall pricing; verify before budgeting. | $0.14 | $0.28 | $0.0028 | $0 | 2026-08-06 | Official |
FAQ
Is this an exact bill predictor?
No. It is an estimate for AI coding workflows. Provider billing, subscriptions, discounts and limits can vary.
Does the calculator call an AI API?
No. All calculations run locally in your browser and no login is required.
Why do coding agents use more tokens than chat?
They repeatedly consume repository context, command output, diffs, plans and review feedback across many turns.