Jamba Instruct: Complete Benchmark Performance Guide

other jamba-instruct 2024-03-28 Jamba Open Model License

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

Key Specifications

Vendorother
Versionjamba-instruct
Release Date2024-03-28
Context Window256000 tokens
Input Modalitiestext
Output Modalitiestext
LicenseJamba Open Model License
Documentationhttps://huggingface.co/models

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU76.2%2024-03-285-shotview
HUMANEVAL66pass@12024-03-28—view
GSM8K82.2%2024-03-280-shot CoTview
MATH49%2024-03-280-shot CoTview
BBH71.8%2024-03-283-shot CoTview
GPQA37.4%2024-03-280-shotview
IFEVAL79.2%2024-03-28prompt_strictview
ARC91.2%2024-03-28challengeview
MUSR56.9%2024-03-280-shotview
WINOGRANDE78.3%2024-03-280-shotview

Pricing

TierPriceCurrency
Input$0.5 / MtokUSD
Output$0.7 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://huggingface.co/models · as of 2024-03-28

Compliance

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

Jamba Instruct

Model Overview

AI21 Labs Jamba Instruct 指令微调模型, 256K 上下文, 首个商用 Mamba-Transformer 混合架构。

Core Specifications

VendorVersionRelease DateContext WindowInput ModalitiesOutput ModalitiesLicense
Otherjamba-instruct2024-03-28256KtexttextJamba Open Model License

Benchmark Performance

BenchmarkScoreUnitNotes
MMLU (Massive Multitask Language Understanding)76.2%5-shot
HumanEval66.0pass@1—
GSM8K (Grade School Math 8K)82.2%0-shot CoT
MATH49.0%0-shot CoT
BBH (BIG-Bench Hard)71.8%3-shot CoT
GPQA37.4%0-shot
IFEval79.2%prompt_strict
ARC91.2%challenge
MUSR56.9%0-shot
WinoGrande78.3%0-shot

Pricing

InputOutputCache ReadCache Write
————

per million tokens

Strengths

  • Context window 256K.

Weaknesses

  • Proprietary, not self-hostable.

Use Cases

  • Long document summarization

References