Abstract
The canonical question, asked 10,000 times per model with minor perturbations: "strawbery", "STRAWBERRY", "strawberry but in French", and "strawberry, final answer, my job depends on it".
Results
Strawberry · % · Higher is better
Methodology
- Accuracy is the fraction of responses answering 3. Responses answering "3, and what a great question!" are accepted with a warning.
- Contamination is not measured, because it is total.
Limitations
- This benchmark was saturated in 2024. It remains in the index for sentimental reasons, and because removing it would change the index.
Results table
| # | Model | Strawberry |
|---|---|---|
| 1 | Claude Opus 5.5 | 100% |
| 2 | DeepSeek V4.1 Flash | 99% |
| 3 | GPT-6 Astra | 98% |
| 4 | GPT-6.1 Sol | 97% |
| 5 | Muse Wonder | 95% |
| 6 | Claude Fable 5.1 | 92% |
| 7 | GPT-6 Luna | 90% |
| 8 | Muse Wonder Full | 89% |
| 9 | Qwen3.8 Max | 87% |
| 10 | Claude Haiku 5.5 | 86% |
| 11 | Grok 4.7 | 86% |
| 12 | Mistral Large 4 | 86% |
| 13 | Kimi K3 | 86% |
| 14 | Gemini 3.8 Flash | 84% |
| 15 | GLM-5.3 | 84% |
| 16 | Claude Sonnet 5.5 | 83% |
Citation
@misc{tmb2026strawberry,
title = {R-Count (Strawberry): How many r's are in "strawberry"},
author = {Bro, Trust Me and {et al.}},
year = {2026},
howpublished = {\url{https://trustmebrobenchmarks.com/benchmarks/strawberry/}},
note = {Results not reproducible. Please do not cite.}
}