Jamba 1.5 Large: Complete Benchmark Performance Guide

other jamba-1-5-large 2024-08-07 Jamba Open Model License

In-depth analysis of Jamba 1.5 Large performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendorother
Versionjamba-1-5-large
Release Date2024-08-07
Context Window256000 tokens
Input Modalitiestext
Output Modalitiestext
LicenseJamba Open Model License
Documentationhttps://huggingface.co/models

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU84.3%2024-08-075-shotview
HUMANEVAL73.7pass@12024-08-07—view
GSM8K83.5%2024-08-070-shot CoTview
MATH60%2024-08-070-shot CoTview
BBH82.8%2024-08-073-shot CoTview
GPQA50.8%2024-08-070-shotview
IFEVAL70.8%2024-08-07prompt_strictview
ARC93.8%2024-08-07challengeview
MUSR51.1%2024-08-070-shotview
WINOGRANDE84.3%2024-08-070-shotview

Pricing

TierPriceCurrency
Input$2 / MtokUSD
Output$8 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://huggingface.co/models · as of 2024-08-07

Compliance

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

Jamba 1.5 Large

Aperçu du modèle

AI21 Labs Jamba 1.5 Large 混合 SSM-Transformer 模型, 256K 上下文, 总参 398B/活跃 94B, 长上下文突出。

Spécifications principales

FournisseurVersionDate de sortieFenêtre de contexteModalités d’entréeModalités de sortieLicence
Otherjamba-1-5-large2024-08-07256KtexttextJamba Open Model License

Performance aux benchmarks

BenchmarkScoreUnitéNotes
MMLU (Massive Multitask Language Understanding)84.3%5-shot
HumanEval73.7pass@1—
GSM8K (Grade School Math 8K)83.5%0-shot CoT
MATH60.0%0-shot CoT
BBH (BIG-Bench Hard)82.8%3-shot CoT
GPQA50.8%0-shot
IFEval70.8%prompt_strict
ARC93.8%challenge
MUSR51.1%0-shot
WinoGrande84.3%0-shot

Tarification

EntréeSortieLecture cacheÉcriture cache
————

par million de jetons

Forces

  • MMLU score 84.3, strong knowledge reasoning.
  • 上下文窗口 256K，支持长文本。
  • 采用 MoE 混合专家架构。

Faiblesses

  • 闭源专有模型，不支持自托管。

Cas d’usage

  • 代码生成与调试
  • 长文档摘要

Références