OpenAI is taking on Mark Zuckerberg\u2019s Meta by launching its own freely available artificial intelligence models.<\/p>\n
The ChatGPT developer has announced two open-weight large language models, which are free to download and can be customised by a user.<\/p>\n
Meta\u2019s Llama models are available on a similar basis, and OpenAI\u2019s move marks a departure from ChatGPT, which is based on a \u201cclosed\u201d model that cannot be customised.<\/p>\n
Sam Altman, OpenAI\u2019s chief executive, said the company was excited to add to a stack of freely available AI models \u201cbased on democratic values \u2026 and for wide benefit\u201d.<\/p>\n
He added: \u201cWe\u2019re excited to make this model, the result of billions of dollars of research, available to the world to get AI into the hands of the most people possible.\u201d<\/p>\n
Zuckerberg, similarly, has said that making models freely available and customisable will \u201censure that more people around the world have access to the benefits and opportunities of AI\u201d and that power is not concentrated in the hands of a small number of companies.<\/p>\n
However, Meta<\/a> has said his company may need to be \u201ccareful\u201d about taking a similar approach with highly advanced AI models.<\/p>\n Sam Altman recently shared a screenshot of what appeared to be the company\u2019s latest AI model, GPT-5. Photograph: Alexander Drago\/Reuters<\/p>\n DeepSeek, a Chinese rival to OpenAI and Meta, has also released powerful models that can be downloaded and customised freely.<\/p>\n OpenAI said the two models, called gpt-oss-120b and gpt-oss-20b-two, outperformed similarly sized models on reasoning tasks, with the larger, 120b model achieving a near-equal performance to its o4-mini model in terms of core reasoning.<\/p>\n Open AI added that during the testing process it had created \u201cmaliciously fine-tuned\u201d versions of the models to simulate biological and cybersecurity threats but found they were \u201cunable to reach high capability levels\u201d.<\/p>\n The existence of freely available, powerful AI models that can be customised by developers has alarmed some experts who have warned that they could be adapted to cause serious harm<\/a>, such as developing bioweapons.<\/p>\n Meta has defined its Llama models as \u201copen source\u201d, meaning they can also be downloaded and customised freely, while their training datasets, architecture and training code are also provided.<\/p>\n However, the Open Source Initiative, a US-based industry organisation, has said the restrictions Meta places on using its models mean it does not qualify as fully open source. OpenAI has used the term \u201copen weight\u201d, effectively a notch down from open source, which means the model can still be fine tuned \u2013 but its workings are not fully transparent to the developer.<\/p>\n OpenAI\u2019s announcement came amid speculation that a new version of the model that underpins ChatGPT could be released soon. On Sunday Altman shared a screenshot<\/a> of what appeared to be the company\u2019s latest AI model, GPT-5.<\/p>\n Meanwhile, Google has outlined its latest step towards artificial general intelligence<\/a> (AGI) with a new model that allows AI systems to interact with a convincing simulation of the real world.<\/p>\n The Genie 3 \u201cworld model\u201d could be used to train robots and autonomous vehicles as they engage with realistic recreations of environments such as warehouses, according to Google<\/a>.<\/p>\n The US technology company\u2019s AI division, Google DeepMind<\/a>, argues that world models are a key step to achieving AGI, a hypothetical level of AI where a system can carry out most tasks on a par with humans \u2013 rather than just individual tasks such as playing chess or translating languages \u2013 and potentially do someone\u2019s job.<\/p>\n DeepMind said such models would play an important role in the development of AI agents, or systems that carry out tasks autonomously<\/a>.<\/p>\n