Best fit
Implementation, review and follow-up tasks on OpenAI's coding models.
Give Codex the implementation and the fast follow-up.
Codex implements, reviews and follows up on OpenAI's coding models. ProPR plans the task, runs it in isolation and delivers a pull request you inspect line by line. Drive it from the terminal with the ProPR CLI.
Approved planContext and scope attached
GPT coding models
Isolated worker
GitHub pull requestCommits, logs, and follow-ups
Your credentials
Task history kept
GitHub merge decision
Agent fit and access
Codex plans and models
ChatGPT Plus at $20 a month includes Codex on the web, in the CLI and IDE, and on iOS. Pro costs $100, $200 or $500 a month, carries no five-hour limit, and the $500 tier adds Ultrafast mode. OpenAI publishes five-hour message estimates for Plus; weekly limits apply on every plan without a published number.
| Plan | Price (USD) | Models | Published limits |
|---|---|---|---|
| ChatGPT Free / Go | $0 / $8 / month | GPT-6 Luna in the desktop app. | OpenAI lists no CLI access and no numbers. |
| ChatGPT Plus | $20 / month | GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Sol, GPT-6 Luna. | Local messages per five hours: 5–45 on GPT-6 Astra, 15–160 on GPT-6.1 Sol, 35–285 on GPT-6.1 Terra, 15–150 on GPT-6 Sol, 350–3,000 on GPT-6 Luna. Cloud tasks draw on the same allowance; weekly limits apply with no published number. |
| ChatGPT Pro | $100 / $200 / $500 / month | Same models; Ultrafast mode on the $500 tier. | No five-hour limit. Weekly limits apply with no published number. |
| ChatGPT Business | $25 / user / month ($20 billed annually) | Same models. | Same per-seat five-hour estimates as Plus. |
| OpenAI API key | Pay-as-you-go per token | The models your key can reach. | Metered billing with no plan cap; no Codex cloud features. |
Every tier with CLI access runs headless: codex exec runs one task and exits, which is how ProPR drives it.
Sources: OpenAI Codex pricing and limits. Verified November 2026.
That subscription runs in ProPR on GitHub issues and pull requests. Label an issue llm-codex-astra or llm-codex-gpt61-sol and ProPR plans the task, runs Codex on your ChatGPT login in an isolated workspace, and opens the PR. You pay OpenAI directly; ProPR adds no token markup. The same label-and-review path is described step by step in the Claude Code on GitHub issues guide, with Codex alongside.
Access
A ChatGPT subscription or an OpenAI API key.
Role in ProPR
Select Codex per phase or per issue. ProPR tracks the task; you iterate in GitHub.
Pick the OpenAI model that fits the phase.
GPT-6 Astra is the default Codex model. GPT-6.1 Sol is the balanced pick, GPT-6 Luna the fast, low-cost one, and GPT-6 Sol rounds out the generation. Set the reasoning level globally, per model, or per issue with a level-* label, from low to ultra.
| Model | Issue label | Context | Notes |
|---|---|---|---|
| GPT-6 Astra (default) | llm-codex-astra | 1.05M | ProPR's default Codex model for implementation, review, and follow-ups. |
| GPT-6.1 Sol | llm-codex-gpt61-sol | 1.05M | A balance of depth and speed for everyday implementation and review. |
| GPT-6 Sol / GPT-6 Luna | llm-codex-gpt6-sol / -gpt6-luna | 1.05M | GPT-6 Luna for fast, low-cost passes and high-volume follow-ups; GPT-6 Sol for a second GPT-6 opinion on the same issue. |
| GPT-5.6 Sol / Terra / Luna | llm-codex-gpt56-sol / -gpt56-terra / -gpt56-luna | 1.05M | The previous generation, pinned for workflows tuned against it. |
| GPT-5.5 / GPT-5.5 Pro | llm-codex-gpt55 / -gpt55-pro | 1M | Complex implementation, refactors, debugging, review, and frontend-heavy work. |
| GPT-5.4 / GPT-5.4 Pro | llm-codex-gpt54 / -gpt54-pro | 1M | Strong general coding work with an established 5.4 configuration. |
| GPT-5.4 Mini / Nano | llm-codex-gpt54-mini / -gpt54-nano | 400K | Inexpensive passes on an established 5.4 configuration. |
| GPT-5.3 Codex / Codex Spark | llm-codex-gpt53-codex / -spark | 400K | Codex-tuned work, code review, or Spark research-preview testing. |
| GPT-5.2, GPT-5 Mini / Nano | llm-codex-gpt52, -gpt5-mini / -gpt5-nano | 400K | Existing configurations pinned to an older model. |
ProPR’s Codex review budget is 272K tokens on every GPT-6 model, whatever the catalog context says.
Model choice
Pick the OpenAI coding model that fits the phase.Phase routing
Route Codex to implementation and fast follow-ups, the phases where its speed pays off.Native goals
Codex runs Goals in its own goal mode, and your corrections reach the running session live.Usage tracking
With Agent Tank, ProPR shows the Codex capacity each call spent.
Set up Codex
Codex setup, models and credentials in the ProPR docs → ProPR runs the official Codex CLI (@openai/codex). Using Codex on its own today? Read the direct Codex comparison.