Claude 2.1: Complete Benchmark Performance Guide

anthropic 2.1 2023-11-21 Proprietary

In-depth analysis of Claude 2.1 performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendoranthropic
Version2.1
Release Date2023-11-21
Context Window200000 tokens
Input Modalitiestext
Output Modalitiestext
LicenseProprietary
Documentationhttps://docs.anthropic.com/claude/docs

Benchmark Performance

Benchmark Score Unit Evaluated At Notes Source
MMLU 60.9 % 2023-11-21 5-shot view
HUMANEVAL 28.1 pass@1 2023-11-21 — view
GSM8K 45.7 % 2023-11-21 0-shot CoT view
MATH 15.5 % 2023-11-21 0-shot CoT view
BBH 63.8 % 2023-11-21 3-shot CoT view
GPQA 26.4 % 2023-11-21 0-shot view
IFEVAL 55 % 2023-11-21 prompt_strict view
ARC 84.8 % 2023-11-21 challenge view
MUSR 37.7 % 2023-11-21 0-shot view
WINOGRANDE 67.4 % 2023-11-21 0-shot view

Pricing

Tier Price Currency
Input$11 / MtokUSD
Output$32 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://www.anthropic.com/pricing · as of 2023-11-21

Compliance

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

Claude 2.1

Model Overview

Anthropic Claude 2.1, 200K 上下文, 相比 2.0 减少 2 倍幻觉率, 改进长文档召回能力。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Anthropic 2.1 2023-11-21 200K text text Proprietary

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 60.9 % 5-shot
HumanEval 28.1 pass@1 —
GSM8K (Grade School Math 8K) 45.7 % 0-shot CoT
MATH 15.5 % 0-shot CoT
BBH (BIG-Bench Hard) 63.8 % 3-shot CoT
GPQA 26.4 % 0-shot
IFEval 55.0 % prompt_strict
ARC 84.8 % challenge
MUSR 37.7 % 0-shot
WinoGrande 67.4 % 0-shot

Pricing

Input Output Cache Read Cache Write
— — — —

per million tokens

Strengths

  • Context window 200K.

Weaknesses

  • HumanEval 28.1, coding weak.
  • Proprietary, not self-hostable.

Use Cases

  • Long document summarization

References