Anthropic launches Claude Opus 4.8: significant gains in AI, autonomous coding and greater system honesty
Anthropic announced the release of its latest artificial intelligence model, the Claude Opus 4.8, marking a significant advancement in autonomous systems capability. The company highlights crucial improvements in several areas, transforming the model into a more effective and reliable collaborator for complex tasks. Esta update aims to optimize users’ interaction with AI, expanding its potential in professional and technical scenarios.
The new model incorporates innovations in autonomous coding, multidisciplinary reasoning and autonomous computer use, in addition to improving intellectual work and autonomous financial analysis. Essas features position the Claude Opus 4.8 as a robust tool for tackling challenges that require high accuracy and information processing capacity. The arrival of Opus 4.8 reflects an ongoing effort to refine the performance and integrity of artificial intelligence.
Aprimoramentos in performance and reliability
Avaliações carried out by experts revealed that Claude Opus 4.8 proves to be a more reliable and accurate model in its judgments when performing action tasks. Anthropic emphasizes that the improvements in honesty have been substantial. Usuários initially reported that Opus 4.8 has a greater propensity to signal uncertainties about its own functioning, avoiding making unsubstantiated statements. Este behavior raises the bar for transparency and security when interacting with AI.
The company’s internal assessments confirm this perception, indicating that Opus 4.8 is approximately four times less likely to allow flaws in its code to go unnoticed, compared to its predecessor. Essa’s error self-detection capability represents a leap in system robustness and reliability. The model, therefore, is designed to operate with greater autonomy and less risk of propagating inaccurate or incorrect information.
More on this story: Meta debuts artificial intelligence agent for coding, intensifying dispute with industry giants
Avaliações Alignment and Prosocial Traits
Results from alignment assessments suggest that Claude Opus 4.8 reaches new heights in measures of prosocial traits. Isso includes greater support for user autonomy and consistent action in the user’s best interest. The model’s architecture was designed to promote more ethical and human-centered interaction, ensuring that its operations are aligned with the user’s goals.
Rates of misaligned behavior such as deception have been significantly reduced in Opus 4.8, showing lower levels than in Opus 4.7. Esses numbers are similar to the preview version of Claude Mythos. Essa consistency in alignment demonstrates Anthropic’s commitment to developing AI models that are not only powerful, but also responsible and safe in their interactions.
Benchmarks and speed optimization
Benchmarks released by Anthropic indicate the superior performance of Claude Opus 4.8 in encoding tests. The model obtained 69.2% in the SWE-Bench Pro, an index that puts it above competitors such as GPT-5.5 and Gemini 3.1 Pro in this and several other benchmarks. Embora o GPT-5.5 maintain the lead in the terminal encoding benchmark, the overall performance of Opus 4.8 is remarkable.
See Also
China’s Shenzhou-23 mission reveals views of Earth from Tiangong station
Manchester City receives third offer from Barcelona for Rodri worth between €60 million and €80 million
Former boxer Prichard Colón dies aged 33 after long battle with brain injury
The fast mode of Claude Opus 4.8 has also been improved to operate at 2.5 times the speed. Adicionalmente, this mode now costs three times less than previous models. Essa speed and cost-effectiveness optimization expands access to advanced AI capabilities to a greater number of developers and enterprises. The Anthropic seeks to balance high performance with operational efficiency.
Novas features for developers
Anthropic is adding important new features to its product line, complementing the release of Claude Opus 4.8. Essas features aim to offer greater flexibility and control to developers using the platform.
Learn more: Microsoft Foundry integrates Anthropic’s Claude Opus 4.6, expanding AI capabilities for enterprises
- Dynamic Job Fluxos (Search Preview):Claude can now complete larger tasks within Claude Code. Ele can schedule work and run hundreds of subagents in parallel in a single session. It is possible to perform source code-scale migrations, spanning hundreds of thousands of lines of code. The feature is available for Claude Code Enterprise, Team and Max plans.
- Controle of effort:No Claude.ai and Cowork, users can choose the level of effort Claude puts into a response. With a lower setting, Claude will respond more quickly and consume rate limits more slowly. Opus 4.8 uses the high effort level by default, which Anthropic claims offers the best balance between quality and user experience.
- Mensagens API:The Mensagens API accepts system inputs within the message matrix, allowing developers to update Claude instructions during task execution.
Disponibilidade and future developments
Claude Opus 4.8 is now available in all regions, with the price for regular use remaining unchanged compared to the previous version, Opus 4.7. The company guarantees that the transition to the new model will be fluid for existing users.
Anthropic continues its work on developing models with the same capabilities as Opus 4.8, but at a lower cost. Além In addition, the company is focused on a new class of models that will be even smarter than the Opus. Medidas security models for the Claude Mythos model are being developed and tested with a small number of organizations. The expectation is that models in the Mythos class will be made available to all customers in the coming weeks.
