Claude Instant 1.2 Release Notes

Anthropic has released Claude Instant 1.2, a faster and lower-priced model designed for tasks such as casual dialogue, text analysis, summarization, and document comprehension. This update incorporates strengths from Claude 2 to enhance performance in technical reasoning and safety.

Performance Gains in Math and Coding

Claude Instant 1.2 demonstrates measurable improvements in mathematical and programming tasks compared to Claude Instant 1.1. On the Codex evaluation, Claude Instant 1.2 achieved a score of 58.7%, an increase from the 52.8% achieved by version 1.1. In the GSM8K benchmark, the model scored 86.7%, compared to 80.9% for the previous version.

Enhanced Reasoning and Capabilities

Claude Instant 1.2 provides more structured and longer responses and exhibits better adherence to formatting instructions. The model also shows improvements in the following areas:

  • Quote extraction
  • Multilingual capabilities
  • Question answering
  • General reasoning

Safety and Reliability Improvements

Claude Instant 1.2 is more resistant to jailbreaks and exhibits fewer hallucinations, as evidenced by Anthropic's automated red-teaming evaluations. These safety improvements ensure the model is more reliable for business applications.

Availability and Access

Claude Instant 1.2 is available to businesses through the Anthropic API.

Sources

Related

  • Dispatch
  • Dispatch
  • Dispatch
  • Dispatch
  • Dispatch