Breaking

OpenAI Codex Hits Capacity Errors Across Multiple Models

July 9, 2026 at 16:22 EDT

Foundation Models
Software Dev & Coding
Infra & Chips

On July 9, 2026, OpenAI's AI coding tool "Codex" began showing a "Selected model is at capacity. Please try a different model." error when users selected several models. OpenAI has acknowledged the issue on its official status page and is working to identify the cause and restore service.

July 9, 2026 · OpenAI Status

Codex Hits a Wall: "Selected model is at capacity"

OpenAI's AI coding tool is throwing elevated errors when users select several high-reasoning models. The cause: inference serving slots maxed out by surging demand — not a context-length limit.

models affected: GPT-5.4 high, GPT-5.5 xHigh, GPT-5.5

2×

prior temporary capacity expansion — at a cost to stability

2nd

major incident — a similar wave hit around June 16, 2026

One Pro user's quota burn

55% of the allowance consumed in just 20 minutes before a reset — how fast a heavy session eats capacity.

55%

consumed
in 20 min

100%

full
allowance

Where it breaks

Trigger

reasoning effort set to xhigh

Cause

serving slots at ceiling — not context length

Impact

long sessions drop mid-task (stream disconnected)

The upside view

The crushing load is read as proof of popularity — demand spikes around new launches like role-specific plugins and Codex Sites.

The frustration

Reliability feels poor even on Pro; interrupted workflows force cumbersome restarts. Rivals like Claude Code and Cursor set the bar for stable concentrated access.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.