AI BriefingClaude Instant 1.2 Released
Key point
Anthropic released Claude Instant 1.2 via API, featuring improved math, coding, and reasoning capabilities along with strengthened safety.
Details
Anthropic has unveiled its latest version, Claude Instant 1.2, via API. This model is faster and cheaper than before, while retaining powerful performance capable of handling various tasks such as conversation, text analysis, summarization, and document comprehension.
Claude Instant 1.2 integrates the strengths of Claude 2 into practical use cases, achieving meaningful advancements in math, coding, reasoning, and safety. In particular, it generates longer, more structured responses and better adheres to formatting instructions. Quote extraction, multilingual capabilities, and question-answering performance have also been improved.
Key benchmark results are as follows:
- Codex: 58.7% (up from 52.8% in the previous 1.1 model)
- GSM8K: 86.7% (up from 80.9% in the previous 1.1 model)
Improvements have also been made in terms of safety. Automated red-teaming evaluations showed reduced hallucinations and increased resistance to jailbreak attacks.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.