Pro $100 (5x) plan is rate-limited as `plus` on Codex Desktop — access_token has plan_type=prolite but the responses API returns X-Codex-Plan-Type=plus

What version of the Codex App are you using (From “About Codex” dialog)?

26.616.41845 (core 0.142.0-alpha.6)

What subscription do you have?

ChatGPT Pro — $100 / 5x tier

What platform is your computer?

Darwin 25.5.0 arm64 arm (macOS 26.5.1)

What issue are you seeing?

My subscription is ChatGPT Pro ($100 "5x" tier). The ChatGPT web Codex shows the
correct Pro usage allowance for this account. But when I use Codex Desktop (ChatGPT
login, calling the /v1 responses API), the server rate-limits me as "plus" and
repeatedly returns 429 usage_limit_reached, even though the account is barely used.

The key inconsistency, on the same request:

• The OAuth access_token my client sends to https://api.openai.com/v1 carries the

claim: chatgpt_plan_type = "prolite" (the Pro 5x tier — correct)

• The rate-limit response header for that same request returns:

X-Codex-Plan-Type = "plus"

Exact error (HTTP 429), PII removed:
{
"type": "error",
"error": {
"type": "usage_limit_reached",
"message": "The usage limit has been reached",
"plan_type": "plus"
},
"status_code": 429
}
Response headers:
X-Codex-Plan-Type: plus
X-Codex-Active-Limit: premium
X-Codex-Primary-Used-Percent: 100 (5-hour window, 300 min)
X-Codex-Secondary-Used-Percent: 42 (weekly window, 10080 min)
X-Codex-Credits-Balance: 0

So OpenAI's own token says "prolite" while OpenAI's own rate limiter treats the
request as "plus".

What steps can reproduce the bug?

1. Be on the ChatGPT Pro $100 (5x) plan, i.e. plan_type "prolite" (mine activated

the day before this started).

1. Sign in to Codex Desktop using ChatGPT login (auth_mode=Chatgpt; originator=

Codex_Desktop; wire_api=responses).

1. Run a few turns with model gpt-5.5.
2. Within a small number of requests the 5-hour "premium" window reports

X-Codex-Primary-Used-Percent: 100 and the API returns 429 usage_limit_reached
with X-Codex-Plan-Type: plus.

1. At the same time, open Codex on the web for the same account — it shows the

correct Pro allowance and is not exhausted.

Context: the account had only ~30 total requests over a week, yet the 5-hour
window is pinned at 100% while the weekly window is only at 42% — consistent with
"plus"-sized 5-hour limits being applied instead of Pro 5x. (Session ID and token/
context usage can be provided privately on request.)

What is the expected behavior?

A Pro $100 (5x) account (plan_type "prolite") should receive Pro 5x Codex usage
limits (5x of Plus) on the Codex Desktop / responses-API path, and the rate-limit
header should reflect the Pro tier — not "plus". Behavior should match what the
ChatGPT web Codex already shows for the same account.

Additional information

• This is purely on the Codex responses-API rate-limiter path. Web Codex shows the

correct Pro quota for the same account, so the subscription/entitlement itself is
correct — only the Codex rate limiter mis-maps the tier.

• Re-authenticating does not help: a fresh access_token still correctly carries

chatgpt_plan_type=prolite, and the API still returns X-Codex-Plan-Type=plus.

• This may be a recognition/sync gap for the newly introduced Pro $100 (5x /

"prolite") tier on the Codex rate limiter. If it is a known propagation delay for
new Pro accounts, please confirm the expected timeline.

• Possibly related to existing "phantom usage limit" reports (capacity healthy on

web/status, but CLI/Desktop still returns usage_limit_reached).