Sonar

Parameters

70B

Published

Feb 11, 2025

Sonar is an AI model developed by Perplexity (United States), first published in February 2025. It works in the language domain, on tasks such as question answering, chat, language modeling/generation and 2 more.

Epoch AI has no training-compute estimate for this model. The model has 70,000,000,000 parameters.

Access: API access. Its weights are not openly released. It is built on top of Llama 3.3 70B. Epoch AI rates the confidence of this record as confident.

Full record

Organization: Perplexity
Country of organization: United States
Domain: Language
Task: Question answering, Chat, Language modeling/generation, Quantitative reasoning, Search
Parameters: 70,000,000,000
Model accessibility: API access
Open weights: No
Base model: Llama 3.3 70B
Epoch confidence: Confident

Benchmark results

01	LiveBench	Global average	46.9
02	Deepresearchbench	Average score	0.38