leftovers

What’s happening with AI in Russia? The artificial intelligence race has as absolute protagonists to the US and China, and it’s surprising that a power like Russia does not seem to be advancing in this type of field. The truth is that he is doing it, but his situation in this area is worrying. Russia does have AI models. Although they are hardly talked about, there are several AI models that are developed by Russian companies and that, above all, are totally oriented towards their citizens. The country does not have access to Western models such as ChatGPT or Claude, but it does have access to these alternatives: GigaChat: is probably the most advanced Russian model. It is developed by the financial institution Sberbank, and It is available via web although to use it it is necessary to have an account in said entity. Yandex Alice: the company that has already offered a search engine in the image and likeness of Google for years also has its own artificial intelligence model, called Alice AI. It is possible to use from its official website in both English and Russian (Spanish is not supported) like a traditional chatbot. MTS AI – one of the largest telecom operators in Russia also has its own model, MTS AImore aimed at companies and developers with its Cotype models that recently support the ability to create agents for enterprise applications. behind. The veto of chips and advanced technology from the US has been one of the factors that has caused these models to be clearly behind the latest frontier models from companies such as OpenAI or Anthropic. He proves it launched three months ago from GigaChat-3.1-Ultra-702B, an open weights model derived from DeepSeek 702B A36B. In the published benchmarks, the performance of this model is at the level of DeepSeek v3 (launched in December 2024) or Qwen3-235B (launched in April 2025). This model is for example available at Hugging Face. China is your natural ally. China and Russia have long maintained a geopolitical alliance that has an impact in various areas. The curious thing is that in this case that impact is being felt less than one would expect. Russian AI models are based on Chinese open models, but at the moment it seems that These are somewhat old versions which they later adapt to Russian. And as happens with Chinese models, which censor certain topics, Russian models apply that same filtering to avoid sensitive issues. Without chips there is no AI. The big problem Russia faces is the same one China faces, but amplified. They don’t have access to advanced Nvidia chips to train their models, so they have looked Shortcuts to bypass trade restrictions and be able to develop those models. Russia, get in line. Sberbank, for example, is trying to get access to Chinese AI chips like the Huawei Ascend 950 for your projects, but there are two problems here. The first is that these chips are aimed at model inference, not their training. The second, that at the moment those who have priority access to these chips are the Chinese companies themselves, which are reserving large quantities for their future projects. ByteDance, Tencent or Alibaba have placed important orders that leave Russia in a complicated situation. Russian chips in the (distant) future. These problems of access to specialized chips could be solved if Russia manages to boost its semiconductor industry. Baikal Electronics has been working on alternatives to x86 chips from Intel and AMD for some time, but also promise develop AI chips in 2029 or 2030. Once again, commercial and technological vetoes mean that in both scenarios the company’s proposals cannot compete with the latest advances of its Western competitors, at least for now. Industrial and military use. Russian LLM developments not only try to present an option for Russian citizens, but also for military applications. The war with Ukraine has actually revealed how Russia is using the well-known AI miniPCs Nvidia Jetson for its Shahed missiles. Russia has it very difficult. The current situation in Russia makes it difficult for the country to present notable alternatives to the most advanced AI models of the US or China (or even Europe). The war with Ukraine also caused an exodus of talent and Russian engineers, although companies like Sber have tried to boost campaigns to attract talent in the university environment. All of this adds to Russia’s difficult access to the most advanced hardware and software and its dependence on an ally like China that is prioritizing its own AI companies. In Xataka | Russia already has its own multi-core CPUs for AI. What it doesn’t have yet is the most important thing: its GPU