Severe Codex Weekly Quota Depletion / Possible Usage Accounting Bug

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

SOL / Luna 5.6 medium

What subscription do you have?

Plus

What platform is your computer?

Windows

What issue are you seeing?

I am experiencing extremely abnormal Codex weekly quota consumption.

Summary

My entire weekly Codex allowance has been exhausted after approximately three development sessions of around 30 minutes each.

That means roughly 1.5 hours of actual development work consumed my entire weekly allowance.

My account now shows that I cannot use Codex again until September 15, 2026.

This is dramatically different from my previous Codex usage and appears to have started recently.

This is NOT limited to GPT-5.6 Sol

Initially I suspected GPT-5.6 Sol because its usage appeared unusually high.

However, I specifically changed my workflow to reduce consumption.

I am using an MCP/workflow that delegates lower-cost work to GPT-5.6 Luna specifically to reduce token and quota consumption.

Despite this, the weekly allowance was still depleted extraordinarily quickly.

Therefore this does not appear to be simply a case of using GPT-5.6 Sol too heavily.

Observed behavior

Approximate usage:

Session 1: ~30 minutes
Session 2: ~30 minutes
Session 3: ~30 minutes

Total active development time:

Approximately 90 minutes

Weekly Codex allowance consumed:

Approximately 100%

Current result:

Codex completely unavailable until September 15, 2026.

The tasks performed were normal software-development tasks and were not unusually large compared with the workloads I have previously used Codex for.

Previously I could use Codex substantially more before approaching the weekly limit.

The effective allowance has therefore decreased by an enormous amount without any corresponding increase in my workload.

Models

I have used:

GPT-5.6 Sol
GPT-5.6 Luna through my token-reduction/delegation workflow

The abnormal consumption persists even when work is delegated to Luna.

Expected behavior

Normal development sessions should consume quota proportionally to the actual compute/token usage.

A few short sessions totaling approximately 1.5 hours should not unexpectedly exhaust an entire weekly allowance when comparable workflows previously lasted substantially longer.

Actual behavior

The weekly quota is currently being consumed at a rate so high that Codex has become practically unusable.

This appears consistent with other recent reports involving unexpectedly depleted Codex weekly quotas and possible quota-accounting/reconciliation problems.

Related reports include:

openai/codex #41969
openai/codex #42660
openai/codex #42765
openai/codex #43351

There are also recent reports of quota decreasing with little or no corresponding visible activity.

Request

Please investigate the server-side Codex usage ledger for my account.

In particular, I would like the team to verify:

Token/compute usage attributed to each Codex session.
Which model generated each usage charge.
Whether delegated/sub-agent/MCP activity is being counted more than once.
Whether context compaction, retries, background activity or failed agent calls are being incorrectly charged.
Whether GPT-5.6 Luna usage is being mapped to the correct quota multiplier.
Whether there has been a recent change to the quota accounting algorithm.
Whether any usage occurred while no active Codex task was running.
Why approximately 90 minutes of normal development exhausted an entire weekly allowance.

This represents a very significant regression compared with my previous Codex usage.

I would appreciate this being escalated to the Codex usage/rate-limit team as a possible quota accounting bug rather than treated as a normal rate-limit complaint.

What steps can reproduce the bug?

I am experiencing extremely abnormal Codex weekly quota consumption.

Summary

My entire weekly Codex allowance has been exhausted after approximately three development sessions of around 30 minutes each.

That means roughly 1.5 hours of actual development work consumed my entire weekly allowance.

My account now shows that I cannot use Codex again until September 15, 2026.

This is dramatically different from my previous Codex usage and appears to have started recently.

This is NOT limited to GPT-5.6 Sol

Initially I suspected GPT-5.6 Sol because its usage appeared unusually high.

However, I specifically changed my workflow to reduce consumption.

I am using an MCP/workflow that delegates lower-cost work to GPT-5.6 Luna specifically to reduce token and quota consumption.

Despite this, the weekly allowance was still depleted extraordinarily quickly.

Therefore this does not appear to be simply a case of using GPT-5.6 Sol too heavily.

Observed behavior

Approximate usage:

Session 1: ~30 minutes
Session 2: ~30 minutes
Session 3: ~30 minutes

Total active development time:

Approximately 90 minutes

Weekly Codex allowance consumed:

Approximately 100%

Current result:

Codex completely unavailable until September 15, 2026.

The tasks performed were normal software-development tasks and were not unusually large compared with the workloads I have previously used Codex for.

Previously I could use Codex substantially more before approaching the weekly limit.

The effective allowance has therefore decreased by an enormous amount without any corresponding increase in my workload.

Models

I have used:

GPT-5.6 Sol
GPT-5.6 Luna through my token-reduction/delegation workflow

The abnormal consumption persists even when work is delegated to Luna.

Expected behavior

Normal development sessions should consume quota proportionally to the actual compute/token usage.

A few short sessions totaling approximately 1.5 hours should not unexpectedly exhaust an entire weekly allowance when comparable workflows previously lasted substantially longer.

Actual behavior

The weekly quota is currently being consumed at a rate so high that Codex has become practically unusable.

This appears consistent with other recent reports involving unexpectedly depleted Codex weekly quotas and possible quota-accounting/reconciliation problems.

Related reports include:

openai/codex #41969
openai/codex #42660
openai/codex #42765
openai/codex #43351

There are also recent reports of quota decreasing with little or no corresponding visible activity.

Request

Please investigate the server-side Codex usage ledger for my account.

In particular, I would like the team to verify:

Token/compute usage attributed to each Codex session.
Which model generated each usage charge.
Whether delegated/sub-agent/MCP activity is being counted more than once.
Whether context compaction, retries, background activity or failed agent calls are being incorrectly charged.
Whether GPT-5.6 Luna usage is being mapped to the correct quota multiplier.
Whether there has been a recent change to the quota accounting algorithm.
Whether any usage occurred while no active Codex task was running.
Why approximately 90 minutes of normal development exhausted an entire weekly allowance.

This represents a very significant regression compared with my previous Codex usage.

I would appreciate this being escalated to the Codex usage/rate-limit team as a possible quota accounting bug rather than treated as a normal rate-limit complaint.

What is the expected behavior?

No response

Additional information

No response

Start by comparing related reports #41969, #42660, #42765, and #43351 with this report’s session and quota details. The investigation needs server-side usage-ledger data for model attribution, duplicate counting, background activity, and quota multipliers; it is complete when the accounting cause is identified and the Codex team can explain or correct the depletion.