TLDR: OpenAI has unveiled two new open-weight GPT models, gpt-oss-120b and gpt-oss-20b, designed to offer high performance with significantly reduced computational requirements. This strategic move empowers telecommunication companies to develop customized, sovereign AI solutions, potentially transforming network management, customer service, and data handling while fostering innovation in the burgeoning sovereign AI market.
OpenAI has made a significant stride in the artificial intelligence landscape with the release of two new open-weight GPT models, gpt-oss-120b and gpt-oss-20b, on August 7, 2025. These models are poised to be a ‘game-changer’ for the telecommunications industry, offering unprecedented flexibility and cost-efficiency for AI deployment.
Unlike fully closed-source models, ‘open-weight’ means that while the training data remains proprietary, the numerical parameters (weights) of the models are disclosed. This allows developers to download, adapt, and deploy the models commercially under the permissive Apache 2.0 license, enabling deep customization for specific tasks and environments. This marks a notable shift in OpenAI’s strategy, responding to market pressures and competition from rivals like Meta’s Llama family and China’s DeepSeek.
A key advantage of these new models is their efficiency. The gpt-oss-20b version, for instance, can run on just 16 GB of memory, making it suitable for edge devices and local inference. The gpt-oss-120b model, while larger, achieves near-parity with OpenAI’s o4-mini on core reasoning benchmarks and runs efficiently on a single 80Gb GPU. Both models demonstrate strong capabilities in reasoning, tool use, few-shot function calling, and CoT reasoning, even outperforming some proprietary models on benchmarks like HealthBench.
For telcos, this development is particularly impactful. As Roy Chua, Founder of AvidThink, explained, ‘The availability of open weight models gives telcos a choice of sovereignty (data never leaves controlled infrastructure), customization (models fine-tuned for specific tasks, computing footprints, regulatory and business needs), and economics (avoiding hyperscaler or AI provider markups while leveraging existing infrastructure investments).’ This enables operators to host AI workloads locally, safeguarding sensitive data and complying with diverse national regulations across their footprints.
Leading telecommunication companies are already embracing this opportunity. Orange Group, for example, has announced it will deploy these advanced models into its infrastructure as an early access partner. Steve Jarrett, Orange’s Chief AI Officer, highlighted that using state-of-the-art open-weight models will drive ‘new use cases to address sensitive enterprise needs, help manage our networks, enable innovating customer care solutions including African regional languages, and much more.’ This initiative also supports digital inclusion and innovation across Africa.
Also Read:
- OpenAI Unveils New Open-Weight Models, Expanding Accessibility and Cloud Partnerships
- OpenAI Unveils Open-Source Models, Integrating with IBM watsonx.ai for Broader Enterprise Adoption
The ability to customize and distill these models for specific use cases, creating smaller sub-models, further enhances their utility for telcos. This strategic move by OpenAI opens up a significant opportunity for the telecommunications sector to tap into the rapidly growing sovereign AI market, which Bank of America’s financial analysts project could reach $50 billion annually.
