[Desktop] Model picker omits GPT-5.6 Sol although the catalog lists it and a direct run succeeds

What version of Codex are you using?

• Codex Desktop: 26.818.31338
• Bundled CLI: 0.149.0-alpha.4
• Platform: macOS 26.6.2, Apple Silicon (arm64)
• Authentication: ChatGPT login (no API key)
• Subscription: individual ChatGPT Pro

What issue are you seeing?

GPT-5.6 Sol is absent from the Codex Desktop model picker.

The picker shows only:

• 5.6 Terra
• 5.6 Luna
• 5.5
• 5.4 Mini

The selected model is 5.6 Terra with High effort. A screenshot is attached.

This does not appear to be a general account-authentication or model-backend availability problem:

1. codex doctor --json reports the Desktop app running, ChatGPT authentication configured, successful app-server initialization, successful provider HTTP reachability, and a successful Responses WebSocket handshake.
2. codex debug models (which refreshes the model catalog by default) lists gpt-5.6-sol as a visible model.
3. A minimal content-free command using the same ChatGPT authentication succeeds when Sol is explicitly selected:

codex exec --model gpt-5.6-sol --sandbox read-only --ephemeral \
  --skip-git-repo-check --color never 'Reply with exactly: OK'

Relevant output:

OpenAI Codex v0.149.0-alpha.4
model: gpt-5.6-sol
provider: openai
approval: never
sandbox: read-only
reasoning effort: high

user
Reply with exactly: OK

codex
OK

Steps to reproduce

1. Sign in to Codex Desktop with an individual ChatGPT Pro account.
2. Start a local Codex task.
3. Open the Model selector.
4. Observe that 5.6 Terra, 5.6 Luna, 5.5, and 5.4 Mini are displayed, but 5.6 Sol is absent.
5. In Terminal, run the minimal command above using the bundled codex executable.
6. Observe that gpt-5.6-sol is accepted and completes successfully.

Expected behavior

The Desktop model picker should present GPT-5.6 Sol when the current model catalog marks it visibility: "list" and the same ChatGPT-authenticated Codex client can successfully start a Sol task. If the model is intentionally withheld from this account or surface, the picker should show a clear model-specific availability explanation instead.

Current catalog evidence

[
  {
    "slug": "gpt-5.6-sol",
    "display_name": "GPT-5.6-Sol",
    "description": "Latest frontier agentic coding model.",
    "visibility": "list",
    "supported_in_api": true,
    "priority": 1,
    "tool_mode": "code_mode_only",
    "multi_agent_version": "v2",
    "context_window": 272000,
    "max_context_window": 872000,
    "supported_reasoning_levels": [
      "low", "medium", "high", "xhigh", "max", "ultra"
    ]
  },
  {
    "slug": "gpt-5.6-terra",
    "display_name": "GPT-5.6-Terra",
    "visibility": "list",
    "supported_in_api": true,
    "priority": 2
  },
  {
    "slug": "gpt-5.6-luna",
    "display_name": "GPT-5.6-Luna",
    "visibility": "list",
    "supported_in_api": true,
    "priority": 3
  }
]

Diagnostic summary

Desktop application: installed, running, version 26.818.31338
Desktop app-server: initialized successfully
Authentication: ChatGPT configured; stored API key: false
Provider reachability: successful
Responses WebSocket: HTTP 101 Switching Protocols
Model provider: OpenAI
Default configured model: gpt-5.6-terra
State database integrity: OK
Network proxy: none detected

Related issues

• #36743 reports Sol missing from the Desktop picker for another ChatGPT Pro account.
• #32520 reports Sol disappearing for a Pro account.

OpenAI’s official model catalog identifies GPT-5.6 Sol as the flagship model for complex reasoning and coding:
https://developers.openai.com/api/docs/models

That documentation describes the API catalog and does not itself establish a ChatGPT plan entitlement. The defect reported here is the inconsistency between Codex Desktop’s picker, its currently supplied catalog, and a successful direct run using the same ChatGPT authentication.

No account email, payment details, auth tokens, local filesystem paths, task content, or session IDs are included.