OpenAI Targets Meta's Llama And DeepSeek's R1 With Laptop-Ready AI Reasoning Tools

Benzinga
2025.08.06 07:08
portai
I'm PortAI, I can summarize articles.

OpenAI has launched two new open-weight language models, GPT-OSS-120B and GPT-OSS-20B, to compete with Meta and DeepSeek. These models can run on laptops and are available through Amazon Web Services' Bedrock platform. OpenAI's CEO, Sam Altman, acknowledged delays in the release and emphasized the need for global feedback. The company is also reportedly raising up to $40 billion in funding, aiming to re-enter the open-model race against competitors.

On Tuesday, OpenAI released two new open-weight language models for the first time since GPT-2, as it aims to challenge rivals like DeepSeek and Meta Platforms, Inc. META with AI tools that can run directly on laptops.

Trending Investment Opportunities

OpenAI Embraces Open Weights With GPT-OSS Models

OpenAI launched two open-weight reasoning models — GPT-OSS-120B and GPT-OSS-20B — marking a major shift in its strategy since the release of GPT-2 in 2019.

Unlike fully open-source models, open-weight models provide access to the trained parameters while keeping the training data and code proprietary.

The GPT-OSS models, trained on a text-only dataset with a focus on math, science and coding, perform comparably to OpenAI's proprietary o3-mini and o4-mini models.

The larger GPT-OSS-120B runs on a single 80 GB GPU, while GPT-OSS-20B is optimized to operate on personal computers.

Amazon Web Services Adds OpenAI Models To Bedrock

Amazon.com, Inc.'s AMZN Web Services also announced that these new models are now available through its Bedrock generative AI marketplace — marking the first time an OpenAI model has been offered on the platform.

In July, Amazon posted net sales of $167.7 billion for the second quarter, marking a 13% increase compared to the same period last year. The figure came in ahead of the Street’s consensus estimate of $161.9 billion.

Amazon Web Services (AWS) contributed $30.9 billion, reflecting a 17.5% year-over-year rise.

Subscribe to the Benzinga Tech Trends newsletter to get all the latest tech developments delivered to your inbox.

Altman Admits Open-Model Delay, Calls For Global Feedback

Earlier this year, OpenAI CEO Sam Altman acknowledged the delay in releasing open-weight models earlier this year, saying, "We've been thinking about this for a long time, but other priorities took precedence."

Altman previously hinted at this shift after China-based DeepSeek launched its R1 model, calling OpenAI's closed approach "on the wrong side of history."

Following the launch, the company will host developer feedback events in San Francisco, Europe and the Asia-Pacific region.

OpenAI, backed by Microsoft Corporation MSFT and valued at $300 billion, is reportedly raising up to $40 billion in a new funding round led by SoftBank Group SFTBFY SFTBY.

With GPT-OSS, OpenAI appears to be re-entering the open-model race against Meta's Llama models and DeepSeek's low-cost, high-performance alternatives.

Photo Courtesy: Primakov on Shutterstock.com

Check out more of Benzinga’s Consumer Tech coverage by following this link.

  • Apple May See Fewer Searches In Safari, But Google CEO Sundar Pichai Insists AI Is Fueling Overall Query Growth: ‘Far From A Zero-Sum Game’