vs
2:0
overall capability wins (0 ties across 2 scored). Sonar leads overall.
- Pricing0:3
- Speed1:1
- Specs2:0
Verdict
Sonar offers ~10.6x larger context (127K vs 12K); Sonar allows ~2.0x more max output (8K vs 4K); Sonar Small is ~5.0x cheaper on a blended token basis than Sonar.
- Sonar offers ~10.6x larger context (127K vs 12K)
- Sonar allows ~2.0x more max output (8K vs 4K)
- Sonar Small is ~5.0x cheaper on a blended token basis than Sonar
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Sonar
- $1.25
- Sonar Small
- $0.25
Sonar Small is about 5.0x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
Perplexity
Perplexity
Family
Sonar
Sonar
License
Proprietary
Proprietary
Open weights
No
No
Release
Jan 27, 2025
Mar 14, 2024
Knowledge cutoff
-
-
API / provider
Perplexity
Perplexity
Modalities
text, image → text
text → text
Specs
Context window
127K
12K
Max output
8K
4K
Parameters
—
—
Pricing
Input $/1M
$1
$0.20
Output $/1M
$1
$0.20
Blended $/1M (3∶1)
$1
$0.20
Speed
tok/s
122
90
TTFT (s)
2.3
1.5
Reasoning
MMLU-Pro
68.9
—
GPQA Diamond
47.1
—
Humanity's Last Exam
4.9
—
AIME 2025
—
—
MATH-500
81.7
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
—
—
AA-LCR v1.1
—
—
CritPt
—
—
MMMU-Pro
—
—
IFBench
—
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
29.5
—
Terminal-Bench 2.1
—
—
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
22.9
—
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
—
—
| Metric | Sonar | Sonar Small | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Perplexity | Perplexity | — |
| Family | Sonar | Sonar | — |
| License | Proprietary | Proprietary | — |
| Open weights | No | No | — |
| Release | Jan 27, 2025 | Mar 14, 2024 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Perplexity | Perplexity | — |
| Modalities | text, image → text | text → text | — |
| Specs | |||
| Context window | 127K | 12K | A +115K |
| Max output | 8K | 4K | A +4K |
| Parameters | — | — | — |
| Pricing | |||
| Input $/1M | $1 | $0.20 | B +$0.80 |
| Output $/1M | $1 | $0.20 | B +$0.80 |
| Blended $/1M (3∶1) | $1 | $0.20 | B +$0.80 |
| Speed | |||
| tok/s | 122 | 90 | A +32 |
| TTFT (s) | 2.3 | 1.5 | B +0.80 |
| Reasoning | |||
| MMLU-Pro | 68.9 | — | — |
| GPQA Diamond | 47.1 | — | — |
| Humanity's Last Exam | 4.9 | — | — |
| AIME 2025 | — | — | — |
| MATH-500 | 81.7 | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | — | — | — |
| AA-LCR v1.1 | — | — | — |
| CritPt | — | — | — |
| MMMU-Pro | — | — | — |
| IFBench | — | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | 29.5 | — | — |
| Terminal-Bench 2.1 | — | — | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 22.9 | — | — |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | — | — | — |
Sonar
- Context
- 127K / 8K out
- Parameters
- —
- Price
- $1 / $1
- Speed
- 122 tok/s · 2.3s TTFT
- Modalities
- text, image → text
- License
- Proprietary
Perplexity's lightweight search-augmented model — real-time web grounding; token pricing excludes per-request search fees ($5–12/1K requests).
Sonar Small
- Context
- 12K / 4K out
- Parameters
- —
- Price
- $0.20 / $0.20
- Speed
- 90 tok/s · 1.5s TTFT
- Modalities
- text → text
- License
- Proprietary
Early Perplexity online SKU (sonar-small-online) — lightweight search-grounded chat before the current Sonar family.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.
These two models have no overlapping published benchmarks in our dataset. Compare specs and pricing instead.