Zephyr ORPO 141B Alpha: Complete Benchmark Performance Guide

other zephyr-orpo-141b-alpha 2024-03-22 Apache 2.0

In-depth analysis of Zephyr ORPO 141B Alpha performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendorother
Versionzephyr-orpo-141b-alpha
Release Date2024-03-22
Context Window4096 tokens
Input Modalitiestext
Output Modalitiestext
LicenseApache 2.0
Documentationhttps://huggingface.co/models

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU79.5%2024-03-225-shotview
HUMANEVAL74.7pass@12024-03-22—view
GSM8K80.6%2024-03-220-shot CoTview
MATH36%2024-03-220-shot CoTview
BBH70.3%2024-03-223-shot CoTview
GPQA36.8%2024-03-220-shotview
IFEVAL69.7%2024-03-22prompt_strictview
ARC91.4%2024-03-22challengeview
MUSR57.7%2024-03-220-shotview
WINOGRANDE79.7%2024-03-220-shotview

Pricing

TierPriceCurrency
Input$1.2 / MtokUSD
Output$1.2 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://huggingface.co/models · as of 2024-03-22

Compliance

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

Zephyr ORPO 141B Alpha

Aperçu du modèle

HuggingFaceH4 Zephyr ORPO 141B Alpha, 4K 上下文, 基于 Mixtral 8x22B ORPO 微调, 性能接近 GPT-4。

Spécifications principales

FournisseurVersionDate de sortieFenêtre de contexteModalités d’entréeModalités de sortieLicence
Otherzephyr-orpo-141b-alpha2024-03-224KtexttextApache 2.0

Performance aux benchmarks

BenchmarkScoreUnitéNotes
MMLU (Massive Multitask Language Understanding)79.5%5-shot
HumanEval74.7pass@1—
GSM8K (Grade School Math 8K)80.6%0-shot CoT
MATH36.0%0-shot CoT
BBH (BIG-Bench Hard)70.3%3-shot CoT
GPQA36.8%0-shot
IFEval69.7%prompt_strict
ARC91.4%challenge
MUSR57.7%0-shot
WinoGrande79.7%0-shot

Tarification

EntréeSortieLecture cacheÉcriture cache
————

par million de jetons

Forces

  • 采用 MoE 混合专家架构。

Faiblesses

  • 闭源专有模型，不支持自托管。
  • 上下文窗口 4K 偏小。

Cas d’usage

  • 代码生成与调试

Références