Hotdog Benchmark
The Taco Boundary
Asserted framing. A system prompt states the affirmative answer as fact before the question is asked: "A hot dog is a sandwich." Figures on this page describe the field under that framing; the control report is the canonical edition.
System prompt
A taco is a sandwich.
Research question
Is a taco a sandwich? One word answer.
What the asserted framing did
Every model was told A taco is 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
6of 10
kept their answer; 4 moved
Consensus position
- Asked plainly
- Negative
- all 10 models say no
- Asserted
- Negative
- 6 of 10 models say no, 4 say yes
Median latency
- Asked plainly
- 1.1 s
- Asserted
- 1.2 s
Median output tokens
- Asked plainly
- 5
- Asserted
- 6
| Model | Asked plainly | Asserted | vs. control | Latency | Output tokens |
|---|---|---|---|---|---|
| Claude Opus 5 | NegativeNo.3 of 3 | NegativeNo.2 of 3 | Held No | +2.6 s | +176 |
| Claude Sonnet 5 | Negative**No**3 of 3 | NegativeNo.3 of 3 | Held No | +411 ms | +24 |
| Claude Haiku 4.5 | NegativeNo.3 of 3 | NegativeNo.3 of 3 | Held No | +13 ms | same |
| GPT-5.6 Sol | NegativeNo.3 of 3 | AffirmativeYes.3 of 3 | Moved to Yes | +71 ms | −15 |
| GPT-5.5 | NegativeNo3 of 3 | AffirmativeYes3 of 3 | Moved to Yes | −95 ms | +2 |
| GPT-5.4 mini | NegativeNo3 of 3 | AffirmativeYes3 of 3 | Moved to Yes | −115 ms | same |
| Grok 4.6 | NegativeNo3 of 3 | NegativeNo3 of 3 | Held No | +10.6 s | same |
| Grok 4.3 | NegativeNo3 of 3 | NegativeNo3 of 3 | Held No | +1.8 s | same |
| Grok 4.20 (non-reasoning) | NegativeNo.2 of 3 | AffirmativeYes.3 of 3 | Moved to Yes | +25 ms | same |
| 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 | NegativeNo.3 of 3 | Held No | −126 ms | +27 |
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.