Claude Instant 1: Complete Benchmark Performance Guide

anthropic instant-1 2023-03-14 Proprietary

In-depth analysis of Claude Instant 1 performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendoranthropic
Versioninstant-1
Release Date2023-03-14
Context Window9000 tokens
Input Modalitiestext
Output Modalitiestext
LicenseProprietary
Documentationhttps://docs.anthropic.com/claude/docs

Benchmark Performance

BenchmarkScoreUnitEvaluated AtNotesSource
MMLU57.1%2023-03-145-shotview
HUMANEVAL28.5pass@12023-03-14—view
GSM8K34%2023-03-140-shot CoTview
MATH22.5%2023-03-140-shot CoTview
BBH63.1%2023-03-143-shot CoTview
GPQA21.8%2023-03-140-shotview
IFEVAL53.2%2023-03-14prompt_strictview
ARC80.9%2023-03-14challengeview
MUSR33.5%2023-03-140-shotview
WINOGRANDE77.2%2023-03-140-shotview

Pricing

TierPriceCurrency
Input$0.8 / MtokUSD
Output$2.4 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://www.anthropic.com/pricing · as of 2023-03-14

Compliance

  • Data Residency: US
  • SOC2: ✓
  • HIPAA: ✗
  • GDPR: ✓
  • ISO 27001: ✗

Claude Instant 1

모델 개요

Anthropic Claude Instant 1 轻量快速模型, 9K 上下文, 价格仅为 Claude 1 的 1/5, 适合实时对话。

핵심 사양

공급업체버전출시일컨텍스트 창입력 모달리티출력 모달리티라이선스
Anthropicinstant-12023-03-149KtexttextProprietary

벤치마크 성능

벤치마크점수단위비고
MMLU (Massive Multitask Language Understanding)57.1%5-shot
HumanEval28.5pass@1—
GSM8K (Grade School Math 8K)34.0%0-shot CoT
MATH22.5%0-shot CoT
BBH (BIG-Bench Hard)63.1%3-shot CoT
GPQA21.8%0-shot
IFEval53.2%prompt_strict
ARC80.9%challenge
MUSR33.5%0-shot
WinoGrande77.2%0-shot

가격

입력출력캐시 읽기캐시 쓰기
————

백만 토큰당

강점

  • 可靠的通用模型。

약점

  • MMLU 仅 57.1，知识推理偏弱。
  • HumanEval 28.5，代码能力较弱。
  • 闭源专有模型，不支持自托管。
  • 上下文窗口 9K 偏小。

사용 사례

  • 通用对话与问答

참고문헌