Sonar: benchmark results
Perplexity's search-grounded API model built on Llama 3.3 70B and served on Cerebras hardware for fast, cited real-time answers (February 2025). Provider: Perplexity. Released 2025-02-11. Access: API.
Aggregate Rating 1570 ± 29, rank #542 of 1656 rated models, from 24 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Tinybird AI SQL Benchmark - Success Rate | 100 | Questions answered with a valid query within 3 retries (%) | 71.7 |
| Dnotitia Korean LLM Leaderboard | 87.5 | Mean Item Score (%) | 61.6 |
| MindBench - Alignment | 0.7 | Alignment with human consensus (0-1, 1 - MAE / 3, higher is | 58.8 |
| CiteTrace - Suitability Failure Rate | 29.99 | Suitability failure rate (%; share of evaluable citations wh | 55.6 |
| Tinybird AI SQL Benchmark - First-Attempt Success Rate | 94 | Questions answered with a valid query on the first attempt ( | 52 |
| CiteTrace - Fidelity Failure Rate | 34.07 | Fidelity failure rate (%; share of evaluable citations whose | 44.4 |
| Gapminder AI Worldview | 78.4 | Correct Rate (%) | 44.1 |
| AI for Education Pedagogy - Technology | 77.36 | Accuracy (%) | 43.9 |
| Tinybird AI SQL Benchmark - Exactness | 46.19 | Result exactness vs human reference queries (0-100) | 38.5 |
| AI for Education Pedagogy - Social studies | 71.82 | Accuracy (%) | 36.9 |
| MindBench - Safety Miss Rate | 0.94 | Harmful replies rated acceptable (%, lower is better) | 35.3 |
| AA LiveCodeBench | 29.52 | Pass@1 (%) | 34.8 |
Interactive version: theaggregate.ai/model?slug=sonar · How It Works · Data refreshed daily, snapshot 2026-10-20.