WizardLM Team WizardLM 2 8x22B: Complete Benchmark Performance Guide

other wizardlmteam-wizardlm-2-8x22b 2024-04-15 Apache 2.0

In-depth analysis of WizardLM Team WizardLM 2 8x22B performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendorother
Versionwizardlmteam-wizardlm-2-8x22b
Release Date2024-04-15
Context Window65536 tokens
Input Modalitiestext
Output Modalitiestext
LicenseApache 2.0
Documentationhttps://huggingface.co/models

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU79.2%2024-04-155-shotview
HUMANEVAL70.2pass@12024-04-15—view
GSM8K77.9%2024-04-150-shot CoTview
MATH37.4%2024-04-150-shot CoTview
BBH80%2024-04-153-shot CoTview
GPQA41.4%2024-04-150-shotview
IFEVAL70.7%2024-04-15prompt_strictview
ARC92.3%2024-04-15challengeview
MUSR56.8%2024-04-150-shotview
WINOGRANDE83.5%2024-04-150-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 2024-04-15

Compliance

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

WizardLM Team WizardLM 2 8x22B

모델 개요

WizardLM Team WizardLM 2 8x22B 官方发布版, 64K 上下文, 基于 Mixtral 8x22B 微调, Arena 上接近 GPT-4。

핵심 사양

공급업체버전출시일컨텍스트 창입력 모달리티출력 모달리티라이선스
Otherwizardlmteam-wizardlm-2-8x22b2024-04-1565KtexttextApache 2.0

벤치마크 성능

벤치마크점수단위비고
MMLU (Massive Multitask Language Understanding)79.2%5-shot
HumanEval70.2pass@1—
GSM8K (Grade School Math 8K)77.9%0-shot CoT
MATH37.4%0-shot CoT
BBH (BIG-Bench Hard)80.0%3-shot CoT
GPQA41.4%0-shot
IFEval70.7%prompt_strict
ARC92.3%challenge
MUSR56.8%0-shot
WinoGrande83.5%0-shot

가격

입력출력캐시 읽기캐시 쓰기
————

백만 토큰당

강점

  • 采用 MoE 混合专家架构。

약점

  • 闭源专有模型，不支持自托管。

사용 사례

  • 代码生成与调试

참고문헌