AI AI Tools Guidebook
Troubleshooting

Cursor Model List Out of Date vs Official Site

Site advertises Opus 4.7 but the local Settings → Models dropdown doesn't show it — old client, stale cache, or staged rollout.

Published: Author: AI Productivity Guide Team 🌐 查看中文版本

The Cursor homepage / changelog headline says “Now supporting Claude Opus 4.7.” You open Cursor → Composer’s model dropdown — opus-4.7 isn’t there; or the list still shows months-old gpt-5.5 and no sign of the new gpt-5-codex. Your account isn’t broken. The usual causes: the local client is out of date, the client-side model-list cache is stale, or the model is rolling out by region / plan and hasn’t reached you yet.

The fix is short, but root-cause matters.

Common causes

1. Cursor client too old

New model support needs matching client code (routing, params, UI labels). Old clients have no idea the model exists. Auto-update occasionally fails; enterprise installs may pin versions.

How to judge: Cursor → About for the version; compare with https://cursor.com/changelog. >2 versions behind = old.

2. Model list cache hasn’t refreshed

Settings → Models pulls a JSON manifest from Cursor’s backend; the client caches it for ~24 hours. Even after an upgrade, the cache can serve a stale list.

How to judge: post-upgrade restart still shows the old list; signing out and back in doesn’t help = cache.

3. Staged rollout by region or plan

New models typically reach Pro / Business first, Hobby a few days later; US nodes first, then APAC; us-east first, then eu-west. For a window you genuinely don’t have it.

How to judge: search forum.cursor.com for “model X rollout”; https://status.cursor.com usually notes rollout state.

4. Plan doesn’t include the model

Some models (opus-4-max-mode, o3-pro, etc.) are Business / Enterprise only; Pro can’t see them. Some are BYOK-only.

How to judge: Settings → Plan + cross-reference https://docs.cursor.com/models.

5. BYOK filters the list down

BYOK paths show only what your provider key actually exposes — Cursor-specific in-house models (cursor-small) won’t appear.

How to judge: Settings → Models — each model is tagged “Cursor” or “Your API Key.”

6. Network restriction / VPN

Some regions or corporate VPNs block Cursor’s models endpoint; the client falls back to a built-in default list.

How to judge: Output → Cursor for “failed to fetch models”; switch network and retry.

Before you start

  • Confirm it’s only your account or everyone in the company; widespread = rollout / plan.
  • Commit settings.json before upgrading so you can tell what a new default changed.
  • Note current Cursor version, plan, network (home / corporate / VPN).

Info to collect

  • Cursor version (Cursor → About) and plan (Settings → Plan).
  • Target model name from docs.cursor.com/models (use the exact model ID).
  • Screenshot of current Settings → Models list.
  • Last 50 lines of View → Output → Cursor (look for fetch errors).
  • BYOK status and whether the provider’s own console shows the model.

Shortest fix path

“Upgrade → flush cache → check rollout.”

Step 1: Upgrade to the latest build

Cursor → Check for Updates (menu bar on macOS; Help menu on Windows / Linux).

If it says “you’re up to date” but changelog shows a release 2 days ago:

# macOS
brew upgrade --cask cursor

# Or download fresh from https://cursor.com/download

For pinned enterprise installs, ping IT.

Step 2: Flush the model-list cache

Option A: Settings → Models → "Refresh" button (top right)
Option B: sign out + sign in (bottom-left avatar)
Option C: nuke the cache:
# macOS
rm -rf ~/Library/Application\ Support/Cursor/Cache
rm -rf ~/Library/Application\ Support/Cursor/Code\ Cache

# Restart Cursor

Step 3: Confirm plan vs availability

Open https://docs.cursor.com/models, check the model’s “Available on” column. Business-only models won’t show up on Pro.

Example matrix (official is source of truth):
- claude-opus-4 (Max mode): Business / Enterprise
- gpt-5-codex: Pro / Business
- o3-pro: Business / Enterprise
- cursor-small: all plans (Cursor proxy only)

Step 4: Check rollout status

https://status.cursor.com for “Model X gradual rollout” notices. forum.cursor.com for “model name rollout.” If it’s rollout, wait — no troubleshooting needed.

Step 5: BYOK to bypass rollout

Settings → Models → API Keys → paste your Anthropic / OpenAI key
→ tick "Use my API key for this model" on the desired entries

Caveat: some Cursor-specific agent integrations may degrade on BYOK paths.

Step 6: Try a different network

If Output shows “failed to fetch models”:

  • Turn off VPN or switch nodes.
  • Corporate proxy may block Cursor’s backend; configure: Cursor → Settings → search “proxy.”

How to verify the fix

  • After upgrade + restart, Settings → Models lists the target model.
  • Composer’s model dropdown shows it as selectable.
  • Send a small prompt against the model; no “model not available” error.

If it still fails

  • Reduce repro to one prompt against one model.
  • Roll back the latest Cursor upgrade to see if a new release dropped the model.
  • Search forum.cursor.com for “model X not showing”; include Cursor version + plan + region.
  • Grab View → Output → Cursor logs and post to Bug Reports — the models team monitors it.

Prevention

  • Manually check for upgrades monthly; don’t trust auto-update alone.
  • Follow https://cursor.com/changelog and forum.cursor.com so you know rollout cadence.
  • When a model isn’t yet available, BYOK as a bridge until Cursor’s path opens.
  • At team level, make sure the plan tier covers the models you actually use.
  • Whitelist Cursor backend domains (cursor.sh, cursor.com) in corporate proxies.

Tags: #Cursor #Troubleshooting

Related Articles