The Register: OpenAI makes good on its name, launches first open weights language models since GPT-2

Aug 5, 2025

—

Source URL: https://www.theregister.com/2025/08/05/openai_open_gpt/
Source: The Register
Title: OpenAI makes good on its name, launches first open weights language models since GPT-2

Feedly Summary: GPT-OSS now available in 120 and 20 billion parameter sizes under Apache 2.0 license
OpenAI released its first open weights language models since GPT-2 on Tuesday with the debut of GPT-OSS.…

AI Summary and Description: Yes

Summary: The release of GPT-OSS, OpenAI’s first open-weight language models since GPT-2, represents a significant advancement in AI and machine learning, highlighting growing trends in open-source AI development. This could have implications for AI security, privacy, and compliance as organizations evaluate the risks and benefits of deploying these models.

Detailed Description: The launch of GPT-OSS under the Apache 2.0 license introduces two new language model sizes—120 billion and 20 billion parameters. This moves OpenAI towards transparency and collaboration within the AI community, generating potential shifts in AI deployment practices and security considerations.

– GPT-OSS is now available in two sizes, targeting a broader audience of developers and businesses.
– The Apache 2.0 licensing allows for more flexible use, encouraging innovation and experimentation.
– This is the first release of open weights language models by OpenAI since GPT-2, which indicates a significant shift in strategy towards open-source and community-driven model development.
– The implications of adopting open-source models include:
– Enhanced ability for organizations to audit and customize models for specific needs.
– Opportunities for improved security practices as developers can identify vulnerabilities more transparently.
– Challenges in ensuring compliance with regulations and standards, especially when customizing and deploying these models in sensitive environments.

Overall, GPT-OSS could catalyze discussions and actions around AI security and governance, as organizations will need to evaluate how to securely integrate these powerful models into their ecosystems while addressing privacy and compliance concerns.

0 license 1 2 2025 5 a Act actions advancement age AGI AI AI development AI security and apach Apache Apache 2 Apache 2.0 Apache 2.0 license as at Audience audit benefits Bi business by C CERN challenge challenges CI CIA co Col collaboration community compliance concerns D day de deployment deployment practices developer developers development drive driven driven model e ecosystem ecosystems end environment exp experimentation first for g Gen GIS Go governance GPT H high Highlight http HTTPS implications in innovation io Iron IRS k l Labor language language model language models learning led Li license licensing low M mac machine Machine Learning Mode model model development models N needs new no o of on ons open open weights open-source open-source models openai OPM opt organization organizations oS oss over parameter per potential Power practices pre privacy pro ps R rag rate RCE re Regulation regulations release Risk risks Ro row s SD sec secure security security considerations security practices sensitive environments shift side Sig size sizes source source model source models specific SSE standards Strategy system systems T the to TP transparency transparent trends two under US use V val vulnerabilities weight Wi x z