AI market news: DeepSeek and Musk on the offensive

Here is the latest information on the valuations of artificial intelligence companies such as OpenAI and Anthropic:​

OpenAI:

• October 2024: OpenAI raised $6.6 billion from investors, bringing the company’s valuation to $157 billion. Investors included Thrive Capital, Khosla Ventures, Microsoft, Nvidia and SoftBank, amongst others. ​
• February 2025: A consortium of investors led by Elon Musk made an offer to acquire OpenAI for $97.4 billion. The offer was rejected by OpenAI’s board of directors.

Anthropic:

• March 2025: Anthropic raised $3.5 billion in a Series E funding round led by Lightspeed Venture Partners, bringing the company’s valuation to $61.5 billion.
• November 2024: Amazon invested a further $4 billion in Anthropic, bringing its total investment to $8 billion. This investment was aimed at strengthening Amazon’s position in the race for dominance in artificial intelligence technology.

It is worth noting that, despite their high valuations, both OpenAI and Anthropic are incurring significant financial losses, mainly due to the high costs associated with developing and maintaining advanced AI models. For example, OpenAI forecasts an operating loss of $5 billion in 2024, despite projected revenues of $3.7 billion. ​​

DeepSeek is a Chinese company specialising in artificial intelligence, founded in 2023 by Liang Wenfeng.The company has developed a range of advanced language models, the latest of which is DeepSeek-V3, launched in December 2024. This model contains 671 billion parameters, of which 37 billion are activated for each token. Thanks to the use of the Mixture-of-Experts (MoE) architecture and the Multi-head Latent Attention (MLA) mechanism, DeepSeek-V3 achieves high performance whilst making efficient use of computational resources.The model was trained on a dataset containing 14.8 trillion tokens, enabling it to outperform other open-source models and match leading proprietary models.

In January 2025, DeepSeek launched the R1 model, which is also based on a MoE architecture with 671 billion parameters, of which 37 billion are activated for each token. The R1 model was designed with a focus on reasoning and problem-solving capabilities, making it competitive with other advanced AI models. ​

Both models, DeepSeek-V3 and R1, are available as open-source software, which enables their widespread use across various fields.

In February 2025, a significant event took place in the technology sector: Elon Musk, together with a consortium of investors, made an offer to acquire OpenAI for $97.4 billion. However, this offer was rejected by OpenAI’s board, causing a major stir in the worlds of technology and business.

Elon Musk and Sam Altman co-founded OpenAI in 2015 as a non-profit organisation aimed at developing artificial intelligence for the benefit of humanity. In 2018, Musk left the company due to differences over its future direction. Since then, OpenAI has transformed into a for-profit organisation, attracting investment from giants such as Microsoft. Musk, criticising this change, set up his own AI company, xAI, and has repeatedly expressed concerns about the direction OpenAI is taking.​​

Musk’s bid highlights the growing tensions within the artificial intelligence sector. His attempt to take over OpenAI could be seen as a bid to regain influence over the direction of AI development. Altman’s rejection of the offer demonstrates OpenAI’s determination to maintain its independence and continue with its chosen business strategy.​

Here is a list of the 10 largest language models (LLMs) currently available, ranked by number of users and capabilities, along with information about their creators:

1. GPT-4o
   Developed by OpenAI, GPT-4o is the ‘omni’ version of GPT-4, unveiled in mid-2024. This model supports multiple modalities, such as text, audio, images and even video, offering advanced capabilities for understanding and generating content in various formats.
2. Claude 3.7 Sonet
   Developed by Anthropic, Claude 3.7 Sonet stands out for its ability to reason transparently and generate content that is similar to that produced by humans, making it one of the leading language models.
3. Gemini 2.0 Flash
   Developed by Google DeepMind, Gemini 2.0 Flash is a multimodal model that integrates advanced language processing capabilities with other modalities, offering versatile applications across a range of fields.
4. Grok 3
   Developed by xAI, Grok 3 is an advanced language model that combines reasoning capabilities with access to platforms such as X (formerly Twitter), enabling more dynamic and up-to-date content generation.
5. LLaMA 3.1
   Developed by Meta AI, LLaMA 3.1 is an open-source language model that offers enhanced reasoning and coding capabilities, outperforming previous versions and rival models in a range of tasks.
6. Falcon 180B
   Developed by the Technology Innovation Institute, Falcon 180B is one of the largest and most powerful open-source language models, designed for tasks such as translation, text generation and research.
7. Qwen-1.5
   Developed by Alibaba Cloud, Qwen-1.5 is an open-source language model that rivals other leading models in terms of capability and cost-effectiveness, designed for high-performance natural language processing tasks.
8. Gemini Ultra
   Developed by Google DeepMind, Gemini Ultra is the first artificial intelligence model to achieve human-level performance on the Massive Multitask Language Understanding (MMLU) test, which measures a model’s ability to perform various language tasks simultaneously.
9. PaLM 2
   Developed by Google, PaLM 2 (Pathways Language Model 2) is an advanced language model that has been used in chatbots such as Bard, offering advanced natural language processing capabilities.
10. Mistral 7B
    Developed by Mistral AI, Mistral 7B is a language model that outperforms GPT-3.5 and Llama 2 70B in many benchmarks, offering advanced natural language processing capabilities with fewer parameters.