[Bug] Thinking mode incorrectly flagged as reasoning extraction during prompt generation benchmark
Status Open
Reported on v2.1.236
Maintainer reply None cached
Activity 0 comments · opened Aug 20, 2026
Bug Description
We are benchmarking different versions of a local llm (qwen3.8 unsloth q4 vs mtplx q8) one of the tests is to run with thinking enabled and with thinking disabled, but for some reason it got flagged as reasoning extraciton, however this test is just running a long prompt generation semvar test.
Environment Info
- Platform: darwin
- Terminal: xterm-256color
- Version: 2.1.236
- Feedback ID: 21ca5629-6b59-413b-b2f8-2ef608885a0f
Errors
[{"error":"Error: ELOOP: too many symbolic links encountered, open '/private/tmp/claude-501/-Users-ismaeltrejo/4c9d9101-03f4-4331-b2d6-5360b81c340f/tasks/brduljc5o.output'\n at async <anonymous> (node:fs/promises:179:56)\n at async #l (/$bunfs/root/cli:439:267)\n at async #c (/$bunfs/root/cli:439:861)","timestamp":"2026-08-20T16:27:05.275Z"},{"error":"Error: ELOOP: too many symbolic links encountered, open '/private/tmp/claude-501/-Users-ismaeltrejo/4c9d9101-03f4-4331-b2d6-5360b81c340f/tasks/bzymxyit7.output'\n at async <anonymous> (node:fs/promises:179:56)\n at async #l (/$bunfs/root/cli:439:267)\n at async #c (/$bunfs/root/cli:439:861)","timestamp":"2026-08-20T16:33:40.730Z"}]