The next leg of the AI race is on , and has expanded beyond the usual players , such asOpenAI , Google , Meta , and Microsoft . In addition to the control of the tech giants , more open - source pick have now taken to the spot with a new focussing in the AI arena .

Various brands , such asDeepSeek , Alibaba , and Baidu , have demonstrated that AI functions can be developed and executed at a fraction of the toll . They have also navigated securing satisfying business partnership and make up one’s mind or continuing to provide AI mathematical product to consumers as destitute or low - price , open source models , while larger company double down on a proprietary , for - earnings trajectory , hide their respectable features behind a paywall .

The previous models , such as DeepSeek ’s R1 , Alibaba ’s Qwen2.5 - VL , and Baidu ’s Ernie 4.5 have been compared to the top models such as OpenAI ’s 4o or GPT-4.5 model . Meanwhile , their codes are easily accessible at depositary , such asGitHubor Huggingface for developers to utilize on alone software .

teach more about these open source AI products and debate if you ’d be interested in making a switch from a more traditional app .

DeepSeek

DeepSeek , unforesightful for Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co. , Ltd. , is a Chinese AI inauguration that advance international attention after its AI framework establish to be formidable in quality to those of well - established brands in the industry , including OpenAI , Google , Meta , Microsoft , and others . The company showcased AI mannikin developed in a fraction of the time and operating price of rival ’ manikin .

The youthful company has been around since April 2023 , an outgrowth of the parent company High - Flyer , but has developed several AI modelling . These include the original DeepSeek - LLM , DeepSeek - Math , DeepSeek V2 , and DeepSeek V3 – the last of which was herald in November 2024 and was already gaining industry attending . The brand announced its DeepSeek - R1 and DeepSeek - R1 - Zero abstract thought models in January 2025 , which lead the AI industry by tempest with comparisons to OpenAI ’s GPT-4 and o1 simulation . The infamous introduction affected stock terms , made competitors rethink business strategy , and make security concerns among regime entities .

DeepSeek was released under the MIT License , which indicates it is as open as it can be without being 100 % overt source . The company telephone its AI example “ subject weight . ” Still , they are available on theGitHub repositoryfor user to download and develop .

DeepSeek is a chatbot - flair AI tool that is available in entanglement - based format as well as through iOS and Android mobile applications . Since gaining ill fame , the companyhas partneredwith several companies , including Lenovo , Tencent , Alibaba , and Baidu , among others , to use its AI models .

Alibaba Qwen

Alibaba Qwen , short for Tongyi Qianwen , a large speech framework family unit , is among the top 10 open - source models in the world . Alibaba itself is a notable Chinese e - commerce sword . The company herald its first model as a genus Beta in April 2023 , which quickly became popular in the AI space as it formally liberate models as open source throughout the twelvemonth .

Subsequent modelshave been loose as either partly or fully opened reference , include Qwen 2 , Qwen - VL2 , Qwen - vl - max , Qwen - Audio , and Qwen2 - Math . The company ’s most recent models include Qwen 2.5 - Max , which was denote in January 2025 , and Qwen2.5 - VL , which was announced in March , market them as more powerful than rival models such as OpenAI ’s GPT-4o , DeepSeek - V3 , and Meta ’s Llama-3.1 - 405 B LLM .

The original Qwen model was base on Meta ’s Llama LLM – another open - source model , modified to suit Alibaba ’s processes . Future model have been turn under the Apache 2.0 License and can be found at theGitHubandHuggingfacerepositories .

Qwen is a chatbot - style AI tool that is available in World Wide Web - based formatting as well as through iOS and Android wandering software . Since its inception , over 90,000 independent models have been train based on Qwen ’s open - source computer code .

Baidu Ernie bot

TheErnie Bot , short for Enhanced Representation through Knowledge Integration , is a chatbot developed by Baidu , a well - known Chinese internet brand . While Baidu has been developing its ERNIE model since 2019 , it first released the Ernie 3.0 LLM in March 2023 .

Since then , it released Ernie 3.5 in June 2023 and Ernie 4.0 in October 2023 . The 4.0 model was set aside for Baidu ’s subscription - base product , while the 3.5 model was used for the brand ’s free chatbot .

presently , the Ernie role model is not open reservoir ; however , the marque plans to soon make it more useable to the public in reception to increase competition in the AI industry , as well as due to rapidly contract production costs .

Baidu released its Ernie 4.5 LLM , and Ernie X1 abstract thought poser in mid - March , which has been compared to OpenAI ’s GPT-4.5 and DeepSeek R1 , respectively . The brand has plans to make its models assailable source on June 30 . to boot , the company announced that its chatbot will be free to all users as of April 1 . A future model , Ernie 5is expected to be released during the 2nd half of 2025 , similarly being open source and devoid to use .

Ernie is a chatbot - trend AI creature that is available in web - based format as well as through iOS and Android mobile applications . Since its inception , Ernie has been used as the simulation for the AI integration on an external version of the Samsung Galaxy S24 series .