Web-grounded search reasoning vs open weights thinking: Sonar Reasoning Pro vs DeepSeek R1.
sonar-reasoning-pro · perplexity
deepseek-reasoner · deepseek
Benchmark
| Workload Scenario | Perplexity Sonar Reasoning Pro | DeepSeek R1 (Reasoner) | Price Delta |
|---|---|---|---|
| 1M input tokens (raw text) | $2.00 | $0.55 | +$1.45 |
| 1M output tokens (generation) | $8.00 | $2.19 | +$5.81 |
| 1M tokens · 70% input / 30% output mix | $3.80 | $1.042 | +$2.758 |
| Standard chat turn (4K in / 800 out, 50% cached) | $0.0144 | $0.003132 | +$0.0113 |
| Monthly scale (10K requests / day) | $4,320.00 | $939.60 | +$3,380.40 |
Negative difference = Perplexity Sonar Reasoning Pro is cheaper. Positive = DeepSeek R1 (Reasoner) is cheaper.
Scaling Curve
Total cost of a token volume at 70/30 input/output split (uncached). Log-log scale.
Verdict
DeepSeek R1 (Reasoner) is cheaper on both input and output rates, so it costs less at every input/output mix. Price alone still isn't the whole decision: capability, latency and context limits (Perplexity Sonar Reasoning Pro: 127.1K, DeepSeek R1 (Reasoner): 64K) may justify the premium for your task.
FAQ
On input, DeepSeek R1 (Reasoner) is cheaper ($0.55/M vs $2.00/M). On output, DeepSeek R1 (Reasoner) is cheaper ($2.19/M vs $8.00/M). The same model is cheaper on both sides, so it wins at every mix.
A chat-style request (4,000 input + 800 output tokens, 50% cached) costs $0.0144 on Perplexity Sonar Reasoning Pro and $0.003132 on DeepSeek R1 (Reasoner) — DeepSeek R1 (Reasoner) is 4.6× more expensive for that workload.
Perplexity Sonar Reasoning Pro supports 127,072 tokens (8.2K max output); DeepSeek R1 (Reasoner) supports 64,000 (8K max output). Perplexity Sonar Reasoning Pro fits 2.0× more context, which matters for long documents and agents.