Model

Sonar

CLOSED

Perplexity AI · Sonar family · released Feb 11, 2025

Web-grounded default search-answer model, fine-tuned from Llama 3.3 70B and served on Cerebras Wafer-Scale Engines at ~1200 tok/s.

ReasoningCodingVisionFunction callingTool useAgentic

1101.8

Elo · rank #190

Parameters

70B

Active params

70B (dense)

Context

128K tokens

Architecture

Dense transformer, fine-tune of Meta Llama 3.3 70B Instruct

License

Proprietary

Languages

—

API price (in/out)

$1 / $1

Modalities

text

Benchmark results

Bar shows position within the tracked field; marker = field best

AIMEMath48.7%#119

best: GPT-5.2 · 100.0%

GPQA DiamondReasoning47.1%#197

best: GPT-5.6 · 94.6%

Humanity's Last ExamReasoning7.3%#102

best: Claude Sonnet 5 · 57.4%

LiveCodeBenchCoding29.5%#152

best: DeepSeek-V4-Pro (Think Max) · 93.5%

MATH-500Math81.7%#65

best: GPT-5 · 99.4%

MMLU-ProKnowledge68.9%#96

best: Claude Fable 5 · 91.5%

SimpleQAKnowledge0.8%#48

best: GPT-4.5 · 62.5%

Run it locally

Closed weights — available via API only. No local deployment.

Input / M tok

Output / M tok

Sonar family

Elo progression across releases

SonarFeb 20251101.8

Resources

DOCSdocs.perplexity.ai/getting-started/models APIAvailable via Perplexity AI API

API price $1/$1 · each benchmark row carries its own source badge (see methodology)