AI在线 AI在线

Llama Is Abandoned! Meta Shifts to Claude, Insider Secrets Revealed

Meta's internal development situation has drawn widespread attention. Recently, an engineer revealed that the company has stopped using its own Llama model and instead opted for Claude Sonnet for code writing. This change seems to indicate Meta's disappointment with Llama, especially after the release of Llama4, which received many negative reviews..

Meta's internal development situation has drawn widespread attention. Recently, an engineer revealed that the company has stopped using its own Llama model and instead opted for Claude Sonnet for code writing. This change seems to indicate Meta's disappointment with Llama, especially after the release of Llama4, which received many negative reviews.

Within 36 hours of Llama4's release, users continuously criticized its functionality, and it later became involved in controversies over large model rankings, damaging its reputation. Now, even Meta's internal teams no longer trust this self-developed model, raising doubts about whether there will be a Llama5 in the future. At the same time, Zuckerberg's newly established "Meta Super Intelligence Lab" (MSL) is developing the next-generation model in parallel with the existing generative AI team, but it has not yet revealed whether it will continue to use the Llama brand name.

Claude

Image source note: The image was generated by AI, and the image licensing service provider is Midjourney

Zuckerberg mentioned in a memo that Llama-4.1 and 4.2 are currently under development. Despite this, a departing employee made sharp criticisms about Meta's cultural issues. He pointed out that the generative AI team at Meta faces a culture of fear and a lack of vision, which have seriously affected the success of the project. Although Meta has invested heavily in the AI field and recruited many excellent talents, these internal issues still plague the company.

After switching to Claude, Meta's engineers found that their code writing capabilities significantly improved. As a new AI programming tool, Claude has gained the favor of many developers. Some Apple developers even said that in a project with 20,000 lines of code, they personally wrote less than 1,000 lines, and most of the work was completed by Claude.

In addition, the company behind Claude, Anthropic, has also achieved significant commercial success, with annualized revenue reaching $4 billion, nearly quadrupling since the beginning of the year. This growth rate shows Claude's strong performance in the market.

For Meta, although it is about to increase its investment in the next generation of Llama, choosing to switch tools beforehand to ensure the efficiency of engineers' work is an admission of Llama's poor performance.

相关资讯

Manus AI Official Website and Social Media Undergo Changes, Chinese Users May Be Affected

Recently, the general AI agent company Manus has laid off some of its operations in China and has moved core technical personnel to its headquarters in Singapore. It is reported that the total number of employees of Manus in China is approximately 120.Manus previously responded to this, stating that the adjustment was based on the company's own operational efficiency, aiming to continue focusing on core business development and improve overall operational efficiency.The signs of this adjustment are also reflected on Manus' official platform.
7/12/2025 6:41:38 PM
AI在线

Anthropic 的 Claude Code 工具存漏洞,导致部分系统“变砖”

Anthropic 最新推出的编码工具 Claude Code 遭遇了一些技术问题。据 GitHub 上的用户报告,该工具的自动更新功能存在漏洞,导致部分工作站出现不稳定甚至无法正常运行的情况。
3/7/2025 3:39:58 PM
远洋

小扎官宣Code Llama重量级更新,新增70B版本,但还有能力限制

功能更强大的 Code Llama 70B 模型来了。今天,Meta 正式发布 Code Llama 70B,这是 Code Llama 系列有史以来最大、性能最好的型号。我们正在开源一个全新的改进版 Code Llama,包括一个更大的 70B 参数模型。编写和编辑代码已成为当今人工智能模型最重要的用途之一。事实证明,编写代码的能力对于人工智能模型更严谨、更合理地处理其他领域的信息也非常重要。我为这一进展感到自豪,并期待着将这些进展纳入 Llama 3 和未来的模型中。Code Llama 70B 提供与之前发布
1/30/2024 2:52:00 PM
机器之心
  • 1