open/ms-ai-architect
1
0
Fork 0
Code Issues Pull requests Releases Wiki Activity Actions

docs(ms-ai-architect): KB-refresh MEDIUM 1/33 — chain-of-thought: reasoning_effort per-modell-nyanse (minimal/none/xhigh varierer) + header-bump

Browse source
This commit is contained in:
Kjell Tore Guttormsen 2026-06-24 11:38:23 +02:00
commit 110604a62d
1 changed files with 3 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

4
skills/ms-ai-advisor/references/prompt-engineering/chain-of-thought-prompting.md
View file

@ -1,6 +1,6 @@
# Chain-of-Thought and Reasoning Prompts
**Last updated:** 2026-02
**Last updated:** 2026-06-24
**Status:** GA
**Category:** Prompt Engineering & LLM Optimization
@ -69,6 +69,8 @@ Disse modellene utfører **innebygd resonnering** automatisk og returnerer både
| **Developer messages** | ✅ | ✅ | ✅ | ✅ |
| **Streaming** | ❌ (o1, o1-preview) | ✅ (limited access for o3) | ✅ | ✅ |
> **Per-modell-nyanser (reasoning_effort):** Verdisettet for GPT-5-serien over er et *superset* — faktisk støtte varierer per modell: `minimal` støttes kun av de opprinnelige GPT-5-modellene (ikke gpt-5.1+); `none` kun fra gpt-5.1 og nyere (gpt-5.1 defaulter til `none`); `xhigh` kun fra gpt-5.1-codex-max og nyere. Modeller før gpt-5.1 defaulter til `medium` og støtter ikke `none`. `reasoning_effort` (low/medium/high) gjelder alle reasoning models unntatt o1-mini.
**Eksempel (GPT-5 med reasoning summary):**
```python
from openai import OpenAI