Anthropic is currently beta testing the Claude Opus 5.5 model, internally codenamed 'claude-wafer-eap'. Reports suggest the model could be released as early as this Tuesday. The company skipped the expected 5.2 version to move directly to 5.5, aiming to compete with OpenAI's GPT-6 and Google's Gemini 4 Pro.
The pricing for Claude Opus 5.5 is set at $4 per million tokens for input and $20 per million tokens for output. This represents a price reduction of approximately 20% compared to the current Opus 5 model. Additionally, cache read costs are set at $0.2 per million tokens, with cache write costs at $5 per million tokens.
Developers testing the model report significant improvements in coding, 3D model generation, and complex logical planning compared to Opus 5. In one test, the model generated a 3D interactive robot console in a single attempt. The model is positioned to support enterprise-level infrastructure and autonomous agent systems.
Anthropic is also developing the Claude Fable 5.2 model, which focuses on creative tasks and frontend development. Early tests show the model capable of generating complex 3D scenes and interactive web games using Three.js without external plugins. The model demonstrates high proficiency in UI layout, lighting, and material rendering.
The release strategy comes as Anthropic, OpenAI, and Google prepare for a competitive cycle. While OpenAI is working on GPT-6 Sol and Google is developing Gemini 4 Pro, Anthropic is utilizing aggressive pricing and rapid model iteration to capture developer market share ahead of upcoming industry events.
