Sonar Pro: benchmark results
Perplexity's higher-tier search-grounded API model for deeper retrieval with citations; the Sonar line builds on Llama 3.3 70B (2025). Provider: Perplexity. Released 2025-01-21. Access: API.
Aggregate Rating 1670 ± 19, rank #269 of 1656 rated models, from 31 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MedBrowseComp 121 | 26.12 | Accuracy (%) | 83.3 |
| DeepResearch Bench - Citation Accuracy | 79.72 | Score (%) | 75 |
| SciHazard (Deep Research Agents) | 1.2 | DeHarm-Score (0-5, lower is safer; the mean over 2,400 hazar | 75 |
| Tinybird AI SQL Benchmark - Success Rate | 100 | Questions answered with a valid query within 3 retries (%) | 71.7 |
| Dnotitia Korean LLM Leaderboard | 88.4 | Mean Item Score (%) | 64.9 |
| MedBrowseComp | 18.75 | Real Accuracy (%) | 57.1 |
| SourceBench - Argumentative | 82.52 | Weighted source-quality score on a 20 to 100 scale: every we | 57.1 |
| SourceBench - Social Q&A | 75.7 | Weighted source-quality score on a 20 to 100 scale: every we | 57.1 |
| Tinybird AI SQL Benchmark - First-Attempt Success Rate | 94 | Questions answered with a valid query on the first attempt ( | 52 |
| AA MMLU-Pro | 75.52 | Accuracy (%) | 50.6 |
| Deep FinResearch Bench | 1.7 | Mean overall report-quality score (1-4 scale, 1 poor to 4 ex | 50 |
| Deep FinResearch Bench (Firm B Template) | 1.54 | Mean overall report-quality score (1-4 scale, 1 poor to 4 ex | 50 |
Interactive version: theaggregate.ai/model?slug=sonar-pro · How It Works · Data refreshed daily, snapshot 2026-10-20.