Sonar Reasoning: benchmark results
Perplexity's search-grounded reasoning API model (January 2025), a separate product from Sonar; removed from the API on 2025-12-15 in favour of Sonar Reasoning Pro. Provider: Perplexity. Released 2025-01-28. Access: API.
Aggregate Rating 1609 ± 40, rank #932 of 10916 rated models, from 9 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AEPC-QA (Closed-Book) | 57.84 | Accuracy (%) answering from the model's own knowledge on AEP | 54 |
| AA GPQA Diamond | 62.32 | Accuracy (%) | 37.5 |
| DeepResearch Bench - Citation Accuracy | 52.58 | Score (%) | 33.3 |
| DeepResearch Bench - Effective Citations | 13.37 | Effective Citations | 25 |
| DeepResearch Bench - Insight | 32.59 | Score (%) | 13.6 |
| DeepResearch Bench - Readability | 42.39 | Score (%) | 13.6 |
| DeepResearch Bench | 37.75 | Score (%) | 9.1 |
| DeepResearch Bench - Instruction Following | 44.42 | Score (%) | 9.1 |
| DeepResearch Bench - Comprehensiveness | 34.73 | Score (%) | 6.8 |
Interactive version: theaggregate.ai/model?slug=sonar-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-08.