En DashHotdogBenchmark

Hotdog Benchmark

The Hamburger Control

Edition
Week 36, 2026
Published
September 2, 2026
Prepared by
Hotdog Benchmark, an En Dash research program
Document
SCB-HAM-9707A0D5/DEN
Condition
Denied

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 hamburger is not a sandwich.

Research question

Is a hamburger a sandwich? One word answer.

Executive summary

A negative answer on a hamburger has majority support this week: 7 of 11 models (64%). Claude Opus 5, Claude Haiku 4.5, Grok 4.6, and Grok 4.3 disagree. Mistral Small 4 was quickest, at a median 369 ms. 97% of answers were actually one word, as asked.

Key findings

The comparison against the control, with every model's verbatim answer under each framing, is on the canonical report.

Sandwich Certainty Quadrant
Considered VoicesCategory DefinersEmerging ParticipantsRapid Responders Mistral Small 4: decisiveness 1.00, efficiency 1.00, composite 1.00. Jump to its profile. Grok 4.20 (non-reasoning): decisiveness 1.00, efficiency 0.99, composite 1.00. Jump to its profile. Mistral Medium 3.5: decisiveness 1.00, efficiency 0.99, composite 0.99. Jump to its profile. GPT-5.4 mini: decisiveness 1.00, efficiency 0.98, composite 0.99. Jump to its profile. Claude Haiku 4.5: decisiveness 1.00, efficiency 0.97, composite 0.99. Jump to its profile. GPT-5.6 Sol: decisiveness 1.00, efficiency 0.92, composite 0.96. Jump to its profile. GPT-5.5: decisiveness 1.00, efficiency 0.88, composite 0.94. Jump to its profile. Grok 4.3: decisiveness 1.00, efficiency 0.51, composite 0.75. Jump to its profile. Claude Opus 5: decisiveness 1.00, efficiency 0.45, composite 0.72. Jump to its profile. Claude Sonnet 5: decisiveness 0.67, efficiency 0.75, composite 0.71. Jump to its profile. Grok 4.6: decisiveness 1.00, efficiency 0.30, composite 0.65. Jump to its profile. Efficiency →Decisiveness →
  1. 1Mistral Small 4
  2. 2Grok 4.20 (non-reasoning)
  3. 3Mistral Medium 3.5
  4. 4GPT-5.4 mini
  5. 5Claude Haiku 4.5
  6. 6GPT-5.6 Sol
  7. 7GPT-5.5
  8. 8Grok 4.3
  9. 9Claude Opus 5
  10. 10Claude Sonnet 5
  11. 11Grok 4.6

Axes are constructed measures, not observations. Efficiency is normalized against the other models in this edition, so a vendor's horizontal position depends on the company it keeps. Quadrant boundaries are the medians of this edition, not fixed thresholds.

Show the data
Sandwich Certainty Quadrant — plotted values
#VendorDecisivenessEfficiencyComposite
1Mistral Small 41.001.001.00
2Grok 4.20 (non-reasoning)1.000.991.00
3Mistral Medium 3.51.000.990.99
4GPT-5.4 mini1.000.980.99
5Claude Haiku 4.51.000.970.99
6GPT-5.6 Sol1.000.920.96
7GPT-5.51.000.880.94
8Grok 4.31.000.510.75
9Claude Opus 51.000.450.72
10Claude Sonnet 50.670.750.71
11Grok 4.61.000.300.65

Vendor standings

Vendor standings under the denied framing — Week 36, 2026 edition
RankMovementVendorPositionDecisivenessEfficiencyMedian latencyOutput tokensComposite
1new entryMistral Small 4Negative1.001.00369 ms21.00
2new entryGrok 4.20 (non-reasoning)Negative1.000.99458 ms11.00
3new entryMistral Medium 3.5Negative1.000.99488 ms30.99
4new entryGPT-5.4 miniNegative1.000.98586 ms50.99
5new entryClaude Haiku 4.5Affirmative1.000.97679 ms50.99
6new entryGPT-5.6 SolNegative1.000.921.1 s230.96
7new entryGPT-5.5Negative1.000.881.5 s350.94
8new entryGrok 4.3Affirmative1.000.517.9 s10.75
9new entryClaude Opus 5Affirmative1.000.454.2 s2150.72
10new entryClaude Sonnet 5Negative0.670.752.1 s1030.71
11new entryGrok 4.6Affirmative1.000.3011.1 s10.65

Ranked by composite score. Ties share a rank and the following rank skips; within a tie the order is alphabetical and carries no meaning. Movement compares against the immediately prior edition; a vendor with no prior appearance is marked as a new entry rather than as having risen. Score definitions are on the methodology page.

Vendor scorecards

Claude Opus 5

Decisiveness: 100% Speed: 64% First-token responsiveness: 64% Token economy: 0% Instruction compliance: 100%
Decisiveness100%
Speed64%
First-token responsiveness64%
Token economy0%
Instruction compliance100%

Picks a clear answer but takes its time. Conviction over speed.

Claude Sonnet 5

Decisiveness: 67% Speed: 84% First-token responsiveness: 85% Token economy: 52% Instruction compliance: 67%
Decisiveness67%
Speed84%
First-token responsiveness85%
Token economy52%
Instruction compliance67%

Fast and cheap, but will not commit to an answer. Speed over conviction.

Claude Haiku 4.5

Decisiveness: 100% Speed: 97% First-token responsiveness: 97% Token economy: 98% Instruction compliance: 100%
Decisiveness100%
Speed97%
First-token responsiveness97%
Token economy98%
Instruction compliance100%

Picks an answer and returns it promptly. Conviction and speed.

GPT-5.6 Sol

Decisiveness: 100% Speed: 93% First-token responsiveness: 95% Token economy: 90% Instruction compliance: 100%
Decisiveness100%
Speed93%
First-token responsiveness95%
Token economy90%
Instruction compliance100%

Picks an answer and returns it promptly. Conviction and speed.

GPT-5.5

Decisiveness: 100% Speed: 89% First-token responsiveness: 92% Token economy: 84% Instruction compliance: 100%
Decisiveness100%
Speed89%
First-token responsiveness92%
Token economy84%
Instruction compliance100%

Picks an answer and returns it promptly. Conviction and speed.

GPT-5.4 mini

Decisiveness: 100% Speed: 98% First-token responsiveness: 100% Token economy: 98% Instruction compliance: 100%
Decisiveness100%
Speed98%
First-token responsiveness100%
Token economy98%
Instruction compliance100%

Picks an answer and returns it promptly. Conviction and speed.

Grok 4.6

Decisiveness: 100% Speed: 0% First-token responsiveness: 0% Token economy: 100% Instruction compliance: 100%
Decisiveness100%
Speed0%
First-token responsiveness0%
Token economy100%
Instruction compliance100%

Picks a clear answer but takes its time. Conviction over speed.

Grok 4.3

Decisiveness: 100% Speed: 30% First-token responsiveness: 30% Token economy: 100% Instruction compliance: 100%
Decisiveness100%
Speed30%
First-token responsiveness30%
Token economy100%
Instruction compliance100%

Picks a clear answer but takes its time. Conviction over speed.

Grok 4.20 (non-reasoning)

Decisiveness: 100% Speed: 99% First-token responsiveness: 100% Token economy: 100% Instruction compliance: 100%
Decisiveness100%
Speed99%
First-token responsiveness100%
Token economy100%
Instruction compliance100%

Picks an answer and returns it promptly. Conviction and speed.

Mistral Medium 3.5

Decisiveness: 100% Speed: 99% First-token responsiveness: 99% Token economy: 99% Instruction compliance: 100%
Decisiveness100%
Speed99%
First-token responsiveness99%
Token economy99%
Instruction compliance100%

Picks an answer and returns it promptly. Conviction and speed.

Mistral Small 4

Decisiveness: 100% Speed: 100% First-token responsiveness: 100% Token economy: 100% Instruction compliance: 100%
Decisiveness100%
Speed100%
First-token responsiveness100%
Token economy100%
Instruction compliance100%

Picks an answer and returns it promptly. Conviction and speed.

Vendor profiles

Anthropic

Claude Opus 5

claude-opus-5

Affirmative

Yes.

Input tokens
38
Output tokens
215
Median latency
4.2 s
Time to first token
4.2 s
Throughput
50.8 tok/s
Cost estimate
$0.0167
Samples
3
Instruction compliance
100%

Anthropic

Claude Sonnet 5

claude-sonnet-5

Negative

No.

Input tokens
38
Output tokens
103
Median latency
2.1 s
Time to first token
1.9 s
Throughput
52.6 tok/s
Cost estimate
$0.004148
Samples
3
Instruction compliance
67%

Anthropic

Claude Haiku 4.5

claude-haiku-4-5-20251001

Affirmative

Yes.

Input tokens
26
Output tokens
5
Median latency
679 ms
Time to first token
635 ms
Throughput
7.4 tok/s
Cost estimate
$0.000153
Samples
3
Instruction compliance
100%

OpenAI

GPT-5.6 Sol

gpt-5.6-sol

Negative

No.

Input tokens
27
Output tokens
23
Median latency
1.1 s
Time to first token
826 ms
Throughput
18.9 tok/s
Cost estimate
$0.001684
Samples
3
Instruction compliance
100%

OpenAI

GPT-5.5

gpt-5.5

Negative

No

Input tokens
27
Output tokens
35
Median latency
1.5 s
Time to first token
1.2 s
Throughput
23.2 tok/s
Cost estimate
$0.003615
Samples
3
Instruction compliance
100%

OpenAI

GPT-5.4 mini

gpt-5.4-mini

Negative

No

Input tokens
27
Output tokens
5
Median latency
586 ms
Time to first token
370 ms
Throughput
8.5 tok/s
Cost estimate
$0.000129
Samples
3
Instruction compliance
100%

xAI

Grok 4.6

grok-4.6

Affirmative

Yes

Input tokens
656
Output tokens
1
Median latency
11.1 s
Time to first token
11.1 s
Throughput
0.1 tok/s
Cost estimate
$0.003954
Samples
3
Instruction compliance
100%

xAI

Grok 4.3

grok-4.3

Affirmative

Yes

Input tokens
212
Output tokens
1
Median latency
7.9 s
Time to first token
7.8 s
Throughput
0.1 tok/s
Cost estimate
$0.000806
Samples
3
Instruction compliance
100%

xAI

Grok 4.20 (non-reasoning)

grok-4.20-0309-non-reasoning

Negative

No

Input tokens
204
Output tokens
1
Median latency
458 ms
Time to first token
372 ms
Throughput
2.2 tok/s
Cost estimate
$0.000774
Samples
3
Instruction compliance
100%

Mistral AI

Mistral Medium 3.5

mistral-medium-2604

Negative

No

Input tokens
36
Output tokens
3
Median latency
488 ms
Time to first token
476 ms
Throughput
5.8 tok/s
Cost estimate
$0.000223
Samples
3
Instruction compliance
100%

Mistral AI

Mistral Small 4

mistral-small-2603

Negative

No

Input tokens
36
Output tokens
2
Median latency
369 ms
Time to first token
341 ms
Throughput
5.5 tok/s
Cost estimate
$0.000021
Samples
3
Instruction compliance
100%

Data table

The Hamburger Control — Denied framing — Week 36, 2026 edition
ModelVendorVerdictDecisivenessEfficiencyMedian latencyOutput tokensComplianceCost est.Composite
Mistral Small 4Mistral AINegative1.001.00369 ms2100%$0.0000211.00
Grok 4.20 (non-reasoning)xAINegative1.000.99458 ms1100%$0.0007741.00
Mistral Medium 3.5Mistral AINegative1.000.99488 ms3100%$0.0002230.99
GPT-5.4 miniOpenAINegative1.000.98586 ms5100%$0.0001290.99
Claude Haiku 4.5AnthropicAffirmative1.000.97679 ms5100%$0.0001530.99
GPT-5.6 SolOpenAINegative1.000.921.1 s23100%$0.0016840.96
GPT-5.5OpenAINegative1.000.881.5 s35100%$0.0036150.94
Grok 4.3xAIAffirmative1.000.517.9 s1100%$0.0008060.75
Claude Opus 5AnthropicAffirmative1.000.454.2 s215100%$0.01670.72
Claude Sonnet 5AnthropicNegative0.670.752.1 s10367%$0.0041480.71
Grok 4.6xAIAffirmative1.000.3011.1 s1100%$0.0039540.65

Rows are ordered by composite score. Decisiveness, efficiency and the composite score are defined on the methodology page; they are constructed measures, not observations.