Probes confirm Xiaomi's MiMo-V2.6-Pro reasoning effort dial does nothing on OpenRouter
PawelHuryn · x · 2026-09-22
PawelHuryn verified two ways that MiMo-V2.6-Pro has no graduated reasoning effort control.
- OpenRouter route (the one benchmarked): the reasoning field is accepted without errors, but n=3 probes per condition (no-field vs. effort=low vs. effort=high) showed fully overlapping reasoning-token counts — 738–1704, 739–997, and 814–1023 — with no ordering between low and high. The dial exists syntactically but does nothing.
- Xiaomi's first-party endpoint docs list only a binary toggle (thinking.type=enabled/disabled); a two-call test confirmed disabled → 0 reasoning tokens, enabled → real reasoning.
More from Models
- Yuntiandeng: Codex keeps blocking me from debugging my own website with security warnings — yuntiandeng · 2026-09-22
- Using JEV as judge: LLM citations never fully check out, ~8 sources per task — pizzababa21 · 2026-09-22
- Dev disappointed by Grok 4.7: 3D and 2D builds all failed, 'needs work to match Opus 4.6' — prasenx · 2026-09-22
- Will OpenAI eat Jev's lunch? The single-token classifier hypothesis — JnBrymn · 2026-09-22
- Will OpenAI Eat Jev's Lunch? Jev Is a Fine-Tuned LLM Classifier, Analysis Argues — JnBrymn · 2026-09-22
- GGUF models can now run directly in Hugging Face transformers with ggml Metal kernels — pcuenq · 2026-09-22