Editorial illustration for OpenAI Launches Two Open-Weight Reasoning Models Under Apache License
OpenAI Releases Open-Weight Reasoning Models for Developers
OpenAI releases gpt-oss-120B and gpt-oss-20B under Apache-2.0-style license
For years, OpenAI held its cards close. Now, the company has finally laid a major hand on the table. With the release of gpt-oss-120B and gpt-oss-20B under an Apache 2.0-style license, the conversation around open-weight AI has been upended.
This is the first time since GPT-2 that OpenAI has contributed serious weights to the public commons , and the timing could not be more telling. Because while OpenAI was deciding to open up, China's open-source ecosystem had already surged ahead. A study from MIT and Hugging Face shows China now slightly leads the U.S.
in global open-model downloads. The wave is real, and it's reshaping the landscape. What does OpenAI's move mean for a field already dominated by DeepSeek, Qwen, and others?
That's the question this article explores.
Finally -- and maybe most symbolically -- OpenAI released gpt-oss-120B and gpt-oss-20B, open-weight MoE reasoning models under an Apache 2.0-style license. Whatever you think of their quality (and early open-source users have been loud about their complaints), this is the first time since GPT-2 that OpenAI has put serious weights into the public commons. China's open-source wave goes mainstream If 2023-24 was about Llama and Mistral, 2025 belongs to China's open-weight ecosystem.
A study from MIT and Hugging Face found that China now slightly leads the U.S. in global open-model downloads, largely thanks to DeepSeek and Alibaba's Qwen family. Highlights: DeepSeek-R1 dropped in January as an open-source reasoning model rivaling OpenAI's o1, with MIT-licensed weights and a family of distilled smaller models.
VentureBeat has followed the story from its release to its cybersecurity impact to performance-tuned R1 variants. Kimi K2 Thinking from Moonshot, a "thinking" open-source model that reasons step-by-step with tools, very much in the o1/R1 mold, and is positioned as the best open reasoning model so far in the world. Z.ai shipped GLM-4.5 and GLM-4.5-Air as "agentic" models, open-sourcing base and hybrid reasoning variants on GitHub.
Baidu's ERNIE 4.5 family arrived as a fully open-sourced, multimodal MoE suite under Apache 2.0, including a 0.3B dense model and visual "Thinking" variants focused on charts, STEM, and tool use. Alibaba's Qwen3 line -- including Qwen3-Coder, large reasoning models, and the Qwen3-VL series released over the summer and fall months of 2025 -- continues to set a high bar for open weights in coding, translation, and multimodal reasoning, leading me to declare this past summer as " VentureBeat has been tracking these shifts, including Chinese math and reasoning models like Light-R1-32B and Weibo's tiny VibeThinker-1.5B, which beat DeepSeek baselines on shoestring training budgets.
OpenAI’s move isn’t charity. It’s a grudging nod to a reality they no longer control. China’s open-weight ecosystem didn’t wait for permission.
DeepSeek, Kimi, Qwen, ERNIE , they built, released, and iterated at a pace that made the old keepers of the flame look like they were guarding museum pieces. Now the museum has thrown its doors open. The 120B and 20B weights are out there, Apache 2.0-style, alongside a torrent of rivals from Beijing to Hangzhou.
The commons just got richer. And the question isn’t which model is “best” , it’s who will build something new on top of all of them. That’s a race worth running.
Related Reading
Browse all OpenAI news →
Policy & Regulation Grammarly faces class-action suit over AI ‘Expert Review’ feature
Policy & Regulation Claude Mythos highlights EU AI safety gaps, says researcher Caroli
Policy & Regulation After ditching AI fitness apps and a Fitbit, I return to Peloton classes
Related Topic OpenAI researcher quits, citing distrust over ad‑driven engagement metrics
LLMs & Generative AI
Related Topic OpenAI launches GPT-Image 1.5 with precise editing for enterprise visuals
LLMs & Generative AI
From Archives Qlik: Nearly half of Indian firms see data quality, governance as AI bottlenecks
From Archives Das says affordable compute and national datasets will cut BFSI cycles to weeks
From Archives OpenAI and Mixpanel breach highlights risk of data aggregation for phishing
Business & Startups
From Archives ChatGPT struggles to report the time in chat, yet can embed the date easily
LLMs & Generative AI
Common Questions Answered
What are the specific details of OpenAI's newly released open-weight reasoning models?
OpenAI has launched two open-weight reasoning models named gpt-oss-120B and gpt-oss-20B under an Apache 2.0-style license. These models represent OpenAI's first significant open-source model release since GPT-2, potentially signaling a new approach to AI model accessibility and transparency.
How does the Apache 2.0-style license impact the usability of these new OpenAI models?
The Apache 2.0-style license provides developers and researchers with broad permissions to use, modify, and distribute the models with minimal restrictions. This permissive licensing approach could potentially reshape how AI technologies are accessed and developed by the broader research and technology communities.
What early feedback has been reported about the gpt-oss-120B and gpt-oss-20B models?
Early open-source users have been vocal with complaints about the models' quality and performance. Despite these criticisms, the release represents a significant moment of OpenAI putting substantial model weights into the public domain, which could spark further innovation and research.
