News

Anthropic Launches Claude 4: Focus on Safety and Reasoning

February 22, 2026 2 min read

Anthropic has launched Claude 4, emphasizing breakthrough reasoning capabilities combined with the company’s signature focus on AI safety and alignment.

Key Improvements

Constitutional AI 2.0: Claude 4 introduces an advanced version of Anthropic’s Constitutional AI approach, providing more nuanced ethical reasoning while reducing over-cautious responses that frustrated users.

400K Context Window: Claude 4 expands its context window to 400K tokens, maintaining its lead in document processing capabilities.

Artifact Execution: Building on Claude 3.5’s Artifacts feature, Claude 4 can now execute code, run simulations, and create interactive applications directly within conversations.

Enterprise Controls: New features allow organizations to customize Claude’s behavior within safety boundaries, addressing a key enterprise concern.

Three Model Tiers

Anthropic introduced three versions:

  • Claude 4 Opus: Most capable model for complex reasoning
  • Claude 4 Sonnet: Balanced performance and cost
  • Claude 4 Haiku: Fast, efficient for simple tasks

Safety Leadership

Anthropic emphasized its continued investment in AI safety, announcing a new $100 million safety research initiative. The company claims Claude 4 exhibits improved resistance to jailbreaks while providing more helpful responses for legitimate use cases.

API Pricing

Claude 4 Sonnet is priced at $15 per million input tokens and $45 per million output tokens. Opus pricing will be announced at general availability.

Availability

Claude 4 Sonnet is available immediately on claude.ai and through the API. Opus will launch in April 2026.