Sonar Reasoning: pricing, context window and benchmarks
Sonar Reasoning is a multimodal model from Perplexity. It offers a 200,000-token context window, costs $11.48 per million input tokens and $62.24 per million output tokens, and averages 76.2 across the index's benchmark suite.
Pricing
Sonar Reasoning costs $11.48 per million input tokens and $62.24 per million output tokens. A blended 3:1 input-to-output mix works out to about $24.17 per million tokens. Cached input is listed at $1.68 per million tokens.
| Item | Price | Notes |
|---|---|---|
| Input | $11.48 | per 1M tokens |
| Output | $62.24 | per 1M tokens |
| Cached input | $1.68 | per 1M tokens |
| Blended 3:1 | $24.17 | estimate per 1M tokens |
| 1M in + 1M out | $73.72 | estimate |
Benchmarks
Scores below are modelled estimates on a 0–100 scale. Use them to compare relative position, not as official vendor results.
| Benchmark | Score | Rank | Percentile | Relative |
|---|---|---|---|---|
| MMLU-Pro | 76.8 | 284 | 53th | |
| GPQA Diamond | 74.6 | 152 | 75th | |
| SWE-bench Verified | 64.6 | 220 | 63th | |
| HumanEval | 79.2 | 261 | 57th | |
| MATH-500 | 80.4 | 202 | 66th | |
| AIME 2025 | 70.9 | 239 | 60th | |
| LiveCodeBench | 71.2 | 166 | 72th | |
| IFEval | 85.1 | 189 | 69th | |
| TruthfulQA | 78.0 | 254 | 58th | |
| HellaSwag | 86.2 | 210 | 65th | |
| MMMU | 71.5 | 260 | 57th | |
| Arena Elo | 1404.5 | 196 | 67th |
What Sonar Reasoning is good for
Sonar Reasoning is tuned toward coding work. It is a reasonable candidate for code generation, review, debugging and repository-scale tasks, especially when paired with a retrieval layer. At $11.48 per million input tokens, it sits in the mid-range of the index, and the 200,000-token context window is enough for large files and multi-file diffs.
Strengths
- Large 200,000-token context window for long documents and agent traces.
- Accepts more than text, including image input where the provider supports it.
- Reasoning-oriented profile, useful for multi-step problems and structured analysis.
- Coding benchmarks are a relative strength within its tier.
Trade-offs
- Proprietary licence: you cannot download the weights or self-host the model.
- Premium input pricing at $11.48 per million tokens.
How it compares
| Model | Provider | Context | Input / 1M | Avg score |
|---|---|---|---|---|
| Llama 3.1 | Meta AI | 200000 | $4.16 | 76.3 |
| Command R+ Instruct | Cohere | 200000 | $4.16 | 76.2 |
| Llama 3 Baseten Instruct | Baseten | 200000 | $3.91 | 76.2 |
| Llama 3 Hyperbolic Instruct | Hyperbolic | 200000 | $3.79 | 76.2 |
| Sonar Reasoning | Perplexity | 200000 | $11.48 | 76.2 |
| o3 Mini | OpenAI | 128000 | $10.68 | 76.2 |
FAQ
How much does Sonar Reasoning cost?
Sonar Reasoning is listed at $11.48 per million input tokens and $62.24 per million output tokens. Cached input is $1.68 per million tokens. A blended 3:1 input-to-output mix is about $24.17 per million tokens.
What is the context window of Sonar Reasoning?
Sonar Reasoning has a context window of 200,000 tokens and can emit up to 4,096 output tokens per request in this index.
Is Sonar Reasoning open weights?
No. Sonar Reasoning is a proprietary model with a proprietary licence, so you cannot download the weights.
How does Sonar Reasoning score on benchmarks?
Sonar Reasoning averages 76.2 across the scored benchmarks in this index, ranking #216 of 600 models. Its strongest relative result is Arena Elo at 1404.5.
What is Sonar Reasoning best for?
Sonar Reasoning is tuned toward coding work. It is a reasonable candidate for code generation, review, debugging and repository-scale tasks, especially when paired with a retrieval layer. At $11.48 per million input tokens, it sits in the mid-range of the index, and the 200,000-token context window is enough for large files and multi-file diffs.