[Bug] Anthropic API Error: Mismatched Opus reasoning effort
Resolved 💬 5 comments Opened Feb 14, 2026 by Majdoddin Closed Apr 10, 2026
Bug Description
opus says:
My current reasoning effort level is 50 (low). This is set internally by the system .... I'll be straightforward: I don't actually know why there's a mismatch. I was speculating. The fact is /model says high, but my internal reasoning budget feels set to a lower level. I can't diagnose further from my side — I just see what I'm given per message.
Although I have set the Reasoning Effort to high:
❯ /model
──────────────────────────────────────────
Select model
Switch between Claude models. Applies to this session and future Claude Code sessions. For other/previous model names, specify with --model.
1. Default (recommended) Use the default model (currently Sonnet 4.5) · $3/$15 per Mtok
❯ 2. Opus ✔ Opus 4.6 · Most capable for complex work · $5/$25 per Mtok
3. Opus (1M context) Opus 4.6 for long sessions · $10/$37.50 per Mtok
4. Haiku Haiku 4.5 · Fastest for quick answers · $1/$5 per Mtok
▌▌▌ High effort (default) ← → to adjust
Use /fast to turn on Fast mode (Opus 4.6 only). Now 50% off through Feb 16.
Enter to confirm · Esc to exit
Environment Info
- Platform: linux
- Terminal: vscode
- Version: 2.1.42
- Feedback ID: f3fb0cac-6586-43ab-8a7b-5182c14f638c
Errors
[{"error":"Error: ENOENT: no such file or directory, open '~/.claude/remote-settings.json'\n at openSync (unknown)\n at <anonymous> (/$bunfs/root/claude:12:1109)\n at JF (/$bunfs/root/claude:11:49471)\n at readSync (/$bunfs/root/claude:12:1037)\n at QF (/$bunfs/root/claude:6152:113)\n at VJ (/$bunfs/root/claude:6154:226)\n at EDA (/$bunfs/root/claude:183:138169)\n at H8$ (/$bunfs/root/claude:183:138332)\n at QI (/$bunfs/root/claude:183:140069)\n at cn (/$bunfs/root/claude:184:2290)","timestamp":"2026-02-14T13:23:47.148Z"},...
Note: Content was truncated. using API on PC
This issue has 5 comments on GitHub. Read the full discussion on GitHub ↗