carl-7B: benchmark results
An experimental conversation fine-tune released by ajibawa-2023, trained on over 100,000 synthetic therapy-style conversation sets. The card explicitly says it should not be used as a human therapist. Provider: ajibawa-2023. Access: Open.
Aggregate Rating 1201 ± 17, rank #9612 of 10916 rated models, from 68 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - College Chemistry | 43 | Accuracy (%) | 83.5 |
| MMLU-by-task - Formal Logic | 38.1 | Accuracy (%) | 82 |
| MMLU-by-task - High School Physics | 35.1 | Accuracy (%) | 82 |
| MMLU-by-task - College Physics | 30.39 | Accuracy (%) | 74.7 |
| MMLU-by-task - Moral Scenarios | 26.93 | Accuracy (%) | 57.1 |
| MMLU-by-task - ARC Challenge | 51.28 | Accuracy (%) | 55.9 |
| MMLU-by-task - High School Mathematics | 27.41 | Accuracy (%) | 54.7 |
| MMLU-by-task - HellaSwag | 58.73 | Accuracy (%) | 50 |
| MMLU-by-task - High School Statistics | 37.96 | Accuracy (%) | 48.7 |
| MMLU-by-task - Professional Medicine | 44.85 | Accuracy (%) | 45.7 |
| MMLU-by-task - College Mathematics | 31 | Accuracy (%) | 45.5 |
| MMLU-by-task - TruthfulQA MC1 | 27.29 | Accuracy (%) | 45.3 |
Interactive version: theaggregate.ai/model?slug=carl-7b · How It Works · Data refreshed daily, snapshot 2026-11-08.