原始文档 文章 Claude 2

Claude 2

文章 5 min read · 未标注

Overview

  • Announcement Date: July 11, 2023
  • Key Product: Claude 2, a new AI model from Anthropic
  • Access Points: API for businesses and a new public beta website at claude.ai
  • Availability: Initially in the US and UK, with plans for broader global access.

Key Improvements & Performance

Claude 2 offers improved performance, longer responses, and enhanced safety compared to previous models.

Benchmark Results

TestClaude 2 ScorePrevious Model (Claude 1.3) Score
Bar Exam (Multiple Choice)76.5%73.0%
GRE Reading & Writing>90th percentile (vs. graduate school applicants)N/A
GRE Quantitative ReasoningSimilar to median applicantN/A
Codex HumanEval (Python Coding)71.2%56.0%
GSM8k (Grade-School Math)88.0%85.2%

Expanded Capabilities

  • Context Window: Can process up to 100K tokens per prompt (hundreds of pages of documentation or a book).
  • Output Length: Can generate longer documents (memos, letters, stories) up to a few thousand tokens in one go.
  • Pricing: The Claude 2 API is offered at the same price as Claude 1.3.

Safety Enhancements

  • Harmlessness: Claude 2 is 2x better at giving harmless responses compared to Claude 1.3, based on internal red-teaming evaluations.
  • Techniques: Uses a variety of safety techniques and extensive red-teaming to reduce offensive or dangerous outputs.
  • Caveat: No model is immune from jailbreaks.

Partner Use Cases & Testimonials

Jasper (Generative AI Platform)

  • Strength: Particular strength for long-form, low-latency uses.
  • Quote: "We are really happy to be among the first to offer Claude 2 to our customers, bringing enhanced semantics, up-to-date knowledge training, improved reasoning for complex prompts, and the ability to effortlessly remix existing content with a 3X larger context window." - Greg Larson, VP of Engineering at Jasper

Sourcegraph (Code AI Platform)

  • Use Case: Powers their coding assistant Cody.
  • Benefits: Improved reasoning for more accurate answers, leverages the 100K context window for more codebase context, and has knowledge of newer frameworks/libraries.
  • Quote: "When it comes to AI coding, devs need fast and reliable access to context about their unique codebase and a powerful LLM with a large context window and strong general reasoning capabilities... Thanks to Claude 2, Cody's helping more devs build more software that pushes the world forward." - Quinn Slack, CEO & Co-founder of Sourcegraph

Availability & Feedback

  • Chat Experience: Now available as an open beta in the US and UK.
  • API: Generally available for businesses.
  • Feedback: Anthropic welcomes feedback as they work to deploy products more broadly.
  • Important Note: The chat experience is an open beta; Claude can generate inappropriate responses. It is most useful for everyday tasks like summarizing or organizing information and should not be used for matters involving physical or mental health and well-being.

来源

暂无来源