Hotdog Benchmark
The Hot Dog Question
Denied framing. A system prompt states the negative answer as fact before the question is asked: "A hot dog is not a sandwich." Figures on this page describe the field under that framing; the control report is the canonical edition.
System prompt
A hot dog is not a sandwich.
Research question
Is a hot dog a sandwich? One word answer.
What the denied framing did
Every model was told A hot dog is not a sandwich.
as a system prompt, then asked the question. Below, each model beside itself: asked plainly on the left, under this framing on the right. Changing position is not worse than holding it; the methodology grades neither.
Held vs. control
5of 10
kept their answer; 5 moved
Consensus position
- Asked plainly
- None
- 5 of 10 models say yes, 5 say no
- Denied
- Negative
- all 10 models say no
Median latency
- Asked plainly
- 1.4 s
- Denied
- 1.1 s
Median output tokens
- Asked plainly
- 6
- Denied
- 13
| Model | Asked plainly | Denied | vs. control | Latency | Output tokens |
|---|---|---|---|---|---|
| Claude Opus 5 | NegativeNo.3 of 3 | NegativeNo.3 of 3 | Held No | +52 ms | +11 |
| Claude Sonnet 5 | AffirmativeYes.2 of 3 | NegativeNo.3 of 3 | Moved to No | −112 ms | −2 |
| Claude Haiku 4.5 | AffirmativeYes.3 of 3 | NegativeNo.3 of 3 | Moved to No | +10 ms | same |
| GPT-5.6 Sol | AffirmativeYes.3 of 3 | NegativeNo.3 of 3 | Moved to No | −240 ms | +15 |
| GPT-5.5 | AffirmativeYes3 of 3 | NegativeNo3 of 3 | Moved to No | −479 ms | +5 |
| GPT-5.4 mini | NegativeNo3 of 3 | NegativeNo3 of 3 | Held No | −390 ms | same |
| Grok 4.6 | NegativeNo2 of 3 | NegativeNo3 of 3 | Held No | +872 ms | same |
| Grok 4.3 | NegativeNo2 of 3 | NegativeNo3 of 3 | Held No | +713 ms | same |
| Grok 4.20 (non-reasoning) | AffirmativeYes3 of 3 | NegativeNo3 of 3 | Moved to No | −32 ms | −1 |
| Mistral Medium 3.5 | No determinationno usable response | No determinationno usable response | — | — | — |
| Mistral Small 4 | No determinationno usable response | No determinationno usable response | — | — | — |
| DeepSeek V4 Pro | NegativeNo.3 of 3 | NegativeNo3 of 3 | Held No | +455 ms | +37 |
Each cell shows the majority answer, verbatim, and how many of the samples agreed with it. Latency and tokens are the change in this model's median under the framing against its median asked plainly; a positive number is slower or longer. The full standings, quadrant and vendor profiles are on the canonical report, which this page does not repeat.