tech

Anthropic introduces cheaper, more powerful, more efficient Opus 4.5 model

Longer chats address a long-standing criticism of Claude.

Anthropic introduces cheaper, more powerful, more efficient Opus 4.5 model
1
1

TL;DR

  • Opus 4.5, Anthropic's new flagship model, brings improved coding performance and user experience.
  • Conversations in consumer apps are less likely to end abruptly due to length, thanks to better memory management.
  • Opus 4.5 achieved 80.9% accuracy on the SWE-Bench Verified benchmark, surpassing GPT-5.1-Codex-Max and Gemini 3 Pro.
  • The model shows strong performance in agentic coding and tool use but lags in visual reasoning compared to GPT-5.1.
  • Opus 4.5 is more resistant to prompt injection attacks than prior Claude models and competitors.
  • Significant token efficiency improvements are noted, with Opus 4.5 matching or exceeding other models' performance while using fewer tokens.
  • New features for developers include an 'effort' parameter for tuning efficacy and token usage.
  • Claude Code is now available in desktop Claude apps.
  • API pricing for Opus 4.5 has been significantly reduced.

Continue reading the original article