Zephyr 7B Beta: Complete Benchmark Performance Guide

other zephyr-7b-beta 2023-10-25 MIT

In-depth analysis of Zephyr 7B Beta performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendorother
Versionzephyr-7b-beta
Release Date2023-10-25
Context Window4096 tokens
Input Modalitiestext
Output Modalitiestext
LicenseMIT
Documentationhttps://huggingface.co/models

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU65.1%2023-10-255-shotview
HUMANEVAL38.4pass@12023-10-25—view
GSM8K50.4%2023-10-250-shot CoTview
MATH31.3%2023-10-250-shot CoTview
BBH50%2023-10-253-shot CoTview
GPQA28.6%2023-10-250-shotview
IFEVAL51.2%2023-10-25prompt_strictview
ARC81.8%2023-10-25challengeview
MUSR32.6%2023-10-250-shotview
WINOGRANDE74%2023-10-250-shotview

Pricing

TierPriceCurrency
Input$0.18 / MtokUSD
Output$0.18 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://huggingface.co/models · as of 2023-10-25

Compliance

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

Zephyr 7B Beta

Ikhtisar Model

HuggingFaceH4 Zephyr 7B Beta 对话模型, 4K 上下文, 基于 Mistral 7B DPO 微调, 性能接近 GPT-3.5。

Spesifikasi Inti

VendorVersiTanggal RilisJendela KonteksModalitas InputModalitas OutputLisensi
Otherzephyr-7b-beta2023-10-254KtexttextMIT

Kinerja Benchmark

BenchmarkSkorSatuanCatatan
MMLU (Massive Multitask Language Understanding)65.1%5-shot
HumanEval38.4pass@1—
GSM8K (Grade School Math 8K)50.4%0-shot CoT
MATH31.3%0-shot CoT
BBH (BIG-Bench Hard)50.0%3-shot CoT
GPQA28.6%0-shot
IFEval51.2%prompt_strict
ARC81.8%challenge
MUSR32.6%0-shot
WinoGrande74.0%0-shot

Harga

InputOutputBaca CacheTulis Cache
————

per juta token

Kelebihan

  • 可靠的通用模型。

Kekurangan

  • HumanEval 38.4，代码能力较弱。
  • 闭源专有模型，不支持自托管。
  • 上下文窗口 4K 偏小。

Kasus Penggunaan

  • 通用对话与问答

Referensi