Key Specifications

Vendorother
Versionsonar-large
Release Date2024-05-13
Context Window128000 tokens
Input Modalitiestext
Output Modalitiestext
LicenseProprietary
Documentationhttps://huggingface.co/models

Benchmark Performance

Benchmark Score Unit Evaluated At Notes Source
MMLU 82.5 % 2024-05-13 5-shot view
HUMANEVAL 80.4 pass@1 2024-05-13 view
GSM8K 84.5 % 2024-05-13 0-shot CoT view
MATH 51.6 % 2024-05-13 0-shot CoT view
BBH 81.6 % 2024-05-13 3-shot CoT view
GPQA 41.6 % 2024-05-13 0-shot view
IFEVAL 74.9 % 2024-05-13 prompt_strict view
ARC 94.1 % 2024-05-13 challenge view
MUSR 65.6 % 2024-05-13 0-shot view
WINOGRANDE 81.1 % 2024-05-13 0-shot view

Pricing

Tier Price Currency
Input$1 / MtokUSD
Output$1 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://huggingface.co/models · as of 2024-05-13

Compliance

  • Data Residency: self-host
  • SOC2: ✗
  • HIPAA: ✗
  • GDPR: ✗
  • ISO 27001: ✗

Sonar Large

Model Overview

Perplexity Sonar Large 在线 RAG 模型, 128K 上下文, 基于 Llama 3 70B 微调, 集成实时搜索。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Other sonar-large 2024-05-13 128K text text Proprietary

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 82.5 % 5-shot
HumanEval 80.4 pass@1
GSM8K (Grade School Math 8K) 84.5 % 0-shot CoT
MATH 51.6 % 0-shot CoT
BBH (BIG-Bench Hard) 81.6 % 3-shot CoT
GPQA 41.6 % 0-shot
IFEval 74.9 % prompt_strict
ARC 94.1 % challenge
MUSR 65.6 % 0-shot
WinoGrande 81.1 % 0-shot

Pricing

Input Output Cache Read Cache Write

per million tokens

Strengths

  • MMLU score 82.5, strong knowledge reasoning.
  • HumanEval 80.4, excellent code generation.
  • Enterprise-grade compliance.

Weaknesses

  • Proprietary, not self-hostable.

Use Cases

  • Code generation and debugging
  • Enterprise customer support

References