Our take
Perplexity is a US-based, SOC 2-compliant host that offers proprietary search-augmented reasoning models with measured throughput data. Its catalogue is narrow — just five variants of its own perplexity-sonar models — but the speed differences between tiers are clearly documented.
Use this for search-augmented reasoning where SOC 2 compliance is required. Pick the base sonar model at $1 per million tokens for budget-conscious search queries, or sonar-pro at 75 tps when throughput matters most. Skip it if you need open-weight models, third-party proprietary models, or a verified EU data-residency endpoint.
- SOC 2 compliance is verified in our data.
- Highest measured throughput among its own offerings — sonar-pro at 75 tps, against 8–70 tps for other variants.
- Clear throughput-latency tradeoff across pricing tiers: base sonar at 70 tps and $1/$1 versus reasoning-pro at 8 tps and $2/$8 versus pro at 75 tps and $3/$15.
- Very narrow model catalogue — only 5 tracked offers, all proprietary perplexity-sonar variants; no open-weight or third-party models.
- No verified EU endpoint; data retention and training practices are also undocumented in our data.
- The reasoning-pro variant shows severe throughput degradation at 8 tps, against 70–75 tps for non-reasoning variants at comparable or lower price points.
Privacy & data handling
- ✓
- Yes
- ✗
- No
- No answer on record
The colour says whether the answer favours you, not whether it is a yes — not training on your prompts earns a green cross, no zero-retention option earns a red one.
Prompt retention: none.