Baichuan2 13B Chat: Complete Benchmark Performance Guide

other baichuan2-13b-chat 2023-09-06 Baichuan 2 License

In-depth analysis of Baichuan2 13B Chat performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendorother
Versionbaichuan2-13b-chat
Release Date2023-09-06
Context Window4096 tokens
Input Modalitiestext
Output Modalitiestext
LicenseBaichuan 2 License
Documentationhttps://huggingface.co/models

Benchmark Performance

Benchmark Score Unit Evaluated At Notes Source
MMLU 58.6 % 2023-09-06 5-shot view
HUMANEVAL 49.6 pass@1 2023-09-06 — view
GSM8K 41 % 2023-09-06 0-shot CoT view
MATH 23.4 % 2023-09-06 0-shot CoT view
BBH 64.2 % 2023-09-06 3-shot CoT view
GPQA 34.7 % 2023-09-06 0-shot view
IFEVAL 48.5 % 2023-09-06 prompt_strict view
ARC 83.2 % 2023-09-06 challenge view
MUSR 37.9 % 2023-09-06 0-shot view
WINOGRANDE 69.5 % 2023-09-06 0-shot view

Pricing

Tier Price Currency
Input$0.3 / MtokUSD
Output$0.3 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://huggingface.co/models · as of 2023-09-06

Compliance

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

Baichuan2 13B Chat

Model Overview

百川智能 Baichuan2 13B Chat 对话模型, 4K 上下文, 中英双语能力突出, 适合中文场景部署。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Other baichuan2-13b-chat 2023-09-06 4K text text Baichuan 2 License

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 58.6 % 5-shot
HumanEval 49.6 pass@1 —
GSM8K (Grade School Math 8K) 41.0 % 0-shot CoT
MATH 23.4 % 0-shot CoT
BBH (BIG-Bench Hard) 64.2 % 3-shot CoT
GPQA 34.7 % 0-shot
IFEval 48.5 % prompt_strict
ARC 83.2 % challenge
MUSR 37.9 % 0-shot
WinoGrande 69.5 % 0-shot

Pricing

Input Output Cache Read Cache Write
— — — —

per million tokens

Strengths

  • Reliable general-purpose model.

Weaknesses

  • MMLU 58.6, weak knowledge reasoning.
  • HumanEval 49.6, coding weak.
  • Proprietary, not self-hostable.
  • Context window 4K is limited.

Use Cases

  • General chat and Q&A

References