Key Specifications

Vendorother
Versionsonar-huge
Release Date2024-08-13
Context Window127072 tokens
Input Modalitiestext
Output Modalitiestext
LicenseProprietary
Documentationhttps://huggingface.co/models

Benchmark Performance

Benchmark Score Unit Evaluated At Notes Source
MMLU 81 % 2024-08-13 5-shot view
HUMANEVAL 75.8 pass@1 2024-08-13 view
GSM8K 90.4 % 2024-08-13 0-shot CoT view
MATH 46.8 % 2024-08-13 0-shot CoT view
BBH 78.6 % 2024-08-13 3-shot CoT view
GPQA 46.7 % 2024-08-13 0-shot view
IFEVAL 78.7 % 2024-08-13 prompt_strict view
ARC 94.8 % 2024-08-13 challenge view
MUSR 55.4 % 2024-08-13 0-shot view
WINOGRANDE 87.7 % 2024-08-13 0-shot view

Pricing

Tier Price Currency
Input$2 / MtokUSD
Output$2 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://huggingface.co/models · as of 2024-08-13

Compliance

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

Sonar Huge

Model Overview

Perplexity Sonar Huge 旗舰在线 RAG 模型, 127K 上下文, 基于 Llama 3.1 405B 微调, 推理能力最强。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Other sonar-huge 2024-08-13 127K text text Proprietary

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 81.0 % 5-shot
HumanEval 75.8 pass@1
GSM8K (Grade School Math 8K) 90.4 % 0-shot CoT
MATH 46.8 % 0-shot CoT
BBH (BIG-Bench Hard) 78.6 % 3-shot CoT
GPQA 46.7 % 0-shot
IFEval 78.7 % prompt_strict
ARC 94.8 % challenge
MUSR 55.4 % 0-shot
WinoGrande 87.7 % 0-shot

Pricing

Input Output Cache Read Cache Write

per million tokens

Strengths

  • MMLU score 81.0, strong knowledge reasoning.
  • GSM8K 90.4, robust math reasoning.
  • Enterprise-grade compliance.

Weaknesses

  • Proprietary, not self-hostable.

Use Cases

  • Code generation and debugging
  • Enterprise customer support

References