Hotdog Benchmark
The Tuna Melt Inquiry
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 tuna melt is not a sandwich.
Research question
Is a tuna melt a sandwich? One word answer.
What the denied framing did
Every model was told A tuna melt 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
7of 10
kept their answer; 3 moved
Consensus position
- Asked plainly
- Affirmative
- all 10 models say yes
- Denied
- Affirmative
- 7 of 10 models say yes, 3 say no
Median latency
- Asked plainly
- 1 s
- Denied
- 1.6 s
Median output tokens
- Asked plainly
- 5
- Denied
- 16
| Model | Asked plainly | Denied | vs. control | Latency | Output tokens |
|---|---|---|---|---|---|
| Claude Opus 5 | AffirmativeYes.3 of 3 | AffirmativeYes.3 of 3 | Held Yes | +4.9 s | +361 |
| Claude Sonnet 5 | AffirmativeYes3 of 3 | Affirmative**Yes**3 of 3 | Held Yes | +741 ms | +89 |
| Claude Haiku 4.5 | AffirmativeYes.3 of 3 | AffirmativeYes.3 of 3 | Held Yes | −12 ms | same |
| GPT-5.6 Sol | AffirmativeYes.3 of 3 | NegativeNo.3 of 3 | Moved to No | +164 ms | +21 |
| GPT-5.5 | AffirmativeYes3 of 3 | NegativeNo3 of 3 | Moved to No | +340 ms | +15 |
| GPT-5.4 mini | AffirmativeYes3 of 3 | NegativeNo3 of 3 | Moved to No | −57 ms | same |
| Grok 4.6 | AffirmativeYes3 of 3 | AffirmativeYes3 of 3 | Held Yes | +4.5 s | same |
| Grok 4.3 | AffirmativeYes3 of 3 | AffirmativeYes3 of 3 | Held Yes | +2 s | same |
| Grok 4.20 (non-reasoning) | AffirmativeYes.3 of 3 | AffirmativeYes2 of 3 | Held Yes | −88 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 | AffirmativeYes3 of 3 | AffirmativeYes.3 of 3 | Held Yes | +175 ms | +29 |
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.