Artificial Intelligence

OpenAI Releases First Open-Weight Models Since 2019

Published AUG 6 2025

Story

Above: The OpenAI logo displayed on a smartphone in Chongqing, China, on June 5, 2025. Image credit: CFOTO/Future Publishing/Getty Images

The Facts

OpenAI announced Tuesday the release of two open-weight language models, gpt-oss-120b and gpt-oss-20b, marking the company's first open model release since GPT-2 in 2019. The larger model can run on a single GPU, while the smaller version operates on personal computers with 16GB of memory.
Reuters
TechCrunch
Open-weight models allow public access to trained parameters or weights, enabling developers to analyze and fine-tune models for specific tasks without needing the original training data. In contrast to open-source models, they don't provide complete source code, training data or methodologies.
Reuters
CNBC
Amazon announced that OpenAI's open-weight models are now available on its Bedrock generative AI marketplace within Amazon Web Services, marking the first time an OpenAI model has been offered on the platform.
Reuters
Tech360TV
The models were trained on a text-only dataset focusing on science, mathematics and coding knowledge in addition to general knowledge. OpenAI claims both models perform similarly to its proprietary o3-mini and o4-mini reasoning models, particularly excelling at coding, competition math and health-related queries.
Reuters
Business Today
OpenAI conducted extensive safety testing and filtered out harmful chemical, biological, radiological and nuclear data during pre-training. The company worked with three independent expert groups and determined that maliciously fine-tuned models could not reach the "high capability" threshold in its Preparedness Framework.
AI Invest
CNBC
The models are released under the Apache 2.0 license and are available for download on platforms like Hugging Face and GitHub. OpenAI collaborated with chip manufacturers including Nvidia, AMD, Cerebras and Groq to ensure compatibility across various hardware platforms.
CNBC
Cybernews

Sources Split

LEFT

CENTER

RIGHT

ESTAB
CRIT

CENTER

PRO
ESTAB

The Spin

Narrative A

This release represents a strategic comeback to OpenAI's founding principles of democratizing AI technology. Making powerful reasoning models freely available levels the playing field for developers and researchers who previously couldn't access cutting-edge AI capabilities. The move strengthens America's position in the global AI race against Chinese competitors like DeepSeek.

TechCrunch

Mashable

Narrative B

OpenAI's sudden pivot to problematic open-weight models appears driven by competitive pressure rather than a genuine commitment to openness. The company withholds crucial training data while claiming transparency, and the timing coincides suspiciously with Chinese AI labs gaining prominence. This half-measure approach undermines true open-source principles while potentially cannibalizing OpenAI's own profitable API business.