# Opus 5 + Claude Voice Released (It’s basically Jarvis) Source: https://www.youtube.com/watch?v=Hv8YGUXZY1g Recap page: https://rapidrecap.app/video/Hv8YGUXZY1g Generated: 2026-07-26T14:29:59.325+00:00 --- ## The Gist Anthropic has released Opus 5 at half the price of Fable 5, alongside new voice modes for iOS and desktop apps that mirror features from OpenAI, highlighting an accelerating race for AI agent dominance. ## Quick Overview Anthropic launched Opus 5 as an everyday model for enterprises and developers at half the cost of Fable 5, while rolling out real-time voice modes across iOS and Codex. Meanwhile, Anthropic and OpenAI are rapidly copying each other's features, including browser updates, multi-tab interfaces, and screen recording integrations, as they converge on universal agent super-apps. **Key Points:** - Anthropic introduced Opus 5, a model priced at $5 per million input tokens and $25 per output tokens, coming close to the frontier intelligence of Fable 5 at half the cost. - Dan Shipper reported that Opus 5 required deleting existing complex workflows because it natively argued with instructions and behaved differently than prior models. - Anthropic added a real-time voice mode to its iOS app that connects directly with tools like Google Calendar, Slack, and Notion. - OpenAI launched voice chat inside Codex and added an iOS simulator feature that lets users build and run Swift apps directly in a side panel. - Claude and Codex have independently converged on nearly identical three-tab interfaces for chat, work, and code tasks. - Codex introduced a record and replay feature that turns user screen recordings into executable automation skills. - Chinese developers are rapidly launching equivalent super-apps and models, including Z Code from the GLM team and Kimi Code from Kimi K3. ![Screenshot at 01:00: The official announcement tweet detailing Opus 5 pricing and performance benchmarks against competing models.](https://ss.rapidrecap.app/screens/Hv8YGUXZY1g/00-01-00.jpg) **Context:** The field of generative artificial intelligence is moving past simple chatbots toward autonomous agents capable of complex knowledge work across multiple applications. Anthropic and OpenAI are competing aggressively to build all-in-one super-apps that integrate voice, browser control, and system execution. ## Detailed Analysis Anthropic released Opus 5 as a high-performance model targeted at knowledge workers and developers for half the price of Fable 5. Early testers like Dan Shipper found that the model required stripping out previous custom engineering workflows because it operated effectively out of the box. Alongside the model release, Anthropic and OpenAI rolled out real-time voice modes and cross-platform desktop integration features. Both companies are racing to consolidate tools into unified super-apps, copying each other's browser tabs, mobile voice capabilities, and screen recording tools. ### Opus 5 Release and Performance Anthropic claims Opus 5 approaches Fable 5 intelligence at half the price. - Opus 5 costs $5 per million input tokens and $25 per output tokens, matching the pricing of prior models like Opus 4.8. - Benchmark comparisons show Opus 5 outperforming Fable 5 in agentic coding, knowledge work, and computer use tasks. - Testers note that Opus 5 struggles with complex multi-step instructions if older custom workflow prompts are retained. ![Screenshot at 01:21: A detailed benchmark comparison card illustrating Opus 5 performance across various agentic tasks.](https://ss.rapidrecap.app/screens/Hv8YGUXZY1g/00-01-21.jpg) ### Claude Voice Mode on iOS Anthropic added real-time voice conversation capabilities to the mobile app. - The voice mode integrates directly with connectors like Gmail, Google Calendar, Google Docs, Slack, and Canva. - Users can speak to Claude while it executes cross-app actions and searches files in real time. - Voice mode can operate using the Opus 5, Sonnet, or Haiku models. ![Screenshot at 05:19: A demonstration of interacting with Opus 5 through the new real-time voice mode interface on iOS.](https://ss.rapidrecap.app/screens/Hv8YGUXZY1g/00-05-19.jpg) ### Codex and OpenAI Voice Updates OpenAI introduced parallel voice features and workspace tools inside Codex. - OpenAI added a voice chat button directly to the Codex desktop application interface. - Voice chat requires users to exit voice mode when requesting URL links because voice generation cannot output hyperlinks natively. - Codex introduced built-in iOS simulator support for building and running mobile applications directly inside a side panel. ![Screenshot at 10:33: A live demonstration of initiating a voice chat session inside the Codex desktop application.](https://ss.rapidrecap.app/screens/Hv8YGUXZY1g/00-10-33.jpg) ### Feature Convergence and Copying Anthropic and OpenAI are rapidly copying each other's platform updates. - Both apps feature identical browser tab management systems that allow multi-page web navigation. - The interface layouts of both platforms have converged onto three primary tabs for chat, work, and code tasks. - Both platforms added record-and-replay skills that translate screen activity recordings into automated agent workflows. ![Screenshot at 18:28: A visual comparison showing identical browser and tab controls across competing platforms.](https://ss.rapidrecap.app/screens/Hv8YGUXZY1g/00-18-28.jpg) ### Chinese Competitors and Business Takeaway International alternatives are emerging rapidly alongside western models. - Chinese platforms like Z Code and Kimi Code are releasing advanced models that mirror western super-app capabilities. - Business operators should focus on selecting functional agent workflows rather than spending excessive time comparing base models. - The ultimate goal for all AI labs is to capture everyday users by embedding advanced multi-tool execution into simple chat interfaces. ![Screenshot at 24:48: A breakdown of competing Chinese super-apps and models matching western capabilities.](https://ss.rapidrecap.app/screens/Hv8YGUXZY1g/00-24-48.jpg)