StarCoder2 15B: Complete Benchmark Performance Guide

other starcoder2-15b 2024-02-28 BigCode Open Model License

In-depth analysis of StarCoder2 15B performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendorother
Versionstarcoder2-15b
Release Date2024-02-28
Context Window16384 tokens
Input Modalitiestext
Output Modalitiestext
LicenseBigCode Open Model License
Documentationhttps://huggingface.co/models

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU60.6%2024-02-285-shotview
HUMANEVAL73.5pass@12024-02-28—view
GSM8K69.6%2024-02-280-shot CoTview
MATH46.2%2024-02-280-shot CoTview
BBH62.8%2024-02-283-shot CoTview
GPQA35.3%2024-02-280-shotview
IFEVAL59.6%2024-02-28prompt_strictview
ARC87.6%2024-02-28challengeview
MUSR51.7%2024-02-280-shotview
WINOGRANDE70.2%2024-02-280-shotview

Pricing

TierPriceCurrency
Input$0.3 / MtokUSD
Output$0.3 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://huggingface.co/models · as of 2024-02-28

Compliance

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

StarCoder2 15B

Model Overview

BigCode StarCoder2 15B 代码专用开源模型, 16K 上下文, 训练于 80+ 编程语言, 代码生成能力突出。

Core Specifications

VendorVersionRelease DateContext WindowInput ModalitiesOutput ModalitiesLicense
Otherstarcoder2-15b2024-02-2816KtexttextBigCode Open Model License

Benchmark Performance

BenchmarkScoreUnitNotes
MMLU (Massive Multitask Language Understanding)60.6%5-shot
HumanEval73.5pass@1—
GSM8K (Grade School Math 8K)69.6%0-shot CoT
MATH46.2%0-shot CoT
BBH (BIG-Bench Hard)62.8%3-shot CoT
GPQA35.3%0-shot
IFEval59.6%prompt_strict
ARC87.6%challenge
MUSR51.7%0-shot
WinoGrande70.2%0-shot

Pricing

InputOutputCache ReadCache Write
————

per million tokens

Strengths

  • Reliable general-purpose model.

Weaknesses

  • Proprietary, not self-hostable.
  • Context window 16K is limited.

Use Cases

  • Code generation and debugging

References