On Thursday, Anthropic officially announced Opus 4.8, its most advanced public model. The new model has been rolled out across all platforms, with pricing remaining consistent with the previous Opus version. Notably, Opus 4.8 was released just 41 days after Opus 4.7—a significantly faster update cycle than usual for Anthropic. This is according to a report by
Techcrunch.com. This rapid update may have been prompted by the lukewarm reception of Opus 4.7 by users. Additionally, recent updates to OpenAI's Codex system and Google's Gemini Flash model have pushed Anthropic to stay competitive. Opus 4.8 not only shows high results in benchmark tests but also demonstrates significant improvements in handling ambiguous data.
Anthropic testers note that the new model identifies suspicious cases faster and is less prone to making unfounded claims. Representatives from Bridgewater Associates acknowledged that Opus 4.8 reports errors in analysis and issues in input data more actively than other models, which increases user confidence in the analysis results.
Alongside the new model, Anthropic launched a new feature called "Dynamic Workflows" in the research phase. This system helps large models like Opus manage complex tasks through hundreds of parallel sub-agents. For example, the collaboration between Claude Code and Opus 4.8 enables the migration of codebases with hundreds of thousands of lines and verification based on existing tests.
The company is also working on the Mythos model, which was paused last month due to cybersecurity concerns. According to Anthropic representatives, once necessary safety measures are completed, Mythos-level models will be available to all customers in the coming weeks. For now, Opus 4.8 remains the company's most powerful and reliable tool.
Read “Zamin” on Telegram!
