Cohere Introduces a Low-Cost AI Model with Fewer Chips
Artificial intelligence company Cohere Inc. has introduced a new AI model that offers impressive performance at a much lower cost than its competitors. Unlike tech giants pouring billions into AI research, Cohere has taken a different approach. By focusing on efficiency and business applications, the company has created a powerful model while spending far less than its rivals.
Building AI with Efficiency
Cohere’s latest large language model, called Command A, was built for under $30 million. In contrast, some competitors spend over $100 million to train a single model. The AI industry is known for its massive infrastructure investments, with companies like OpenAI and Anthropic believing that artificial general intelligence (AGI) requires more computational power. However, Cohere proves a targeted and efficient approach can achieve similar results.
Cohere co-founder Nick Frosst emphasized that the company’s success comes from focusing on business needs rather than trying to build AI that can do everything. By training the model specifically for tasks like coding, customer support, and answering technical questions, they have created a cost-effective and high-performing AI.
Instead of building technology that aims to be all-powerful, Cohere trains its models to serve business customers efficiently. This focused approach allows them to use significantly fewer resources than competitors while still delivering competitive AI solutions.
Why AI Companies Are Spending Big on Power
While Cohere is focused on efficiency, other AI companies are in a race for computational dominance. OpenAI, Anthropic, and Elon Musk’s xAI are investing heavily in data centers, GPUs, and AI supercomputers. For example, OpenAI and Oracle plan to invest $500 billion in a massive AI supercomputer project called Stargate. Similarly, xAI is constructing a facility with 100,000 GPUs, with plans to double that number.
On the other hand, Cohere has access to around 8,500 GPUs, a far smaller number than these AI giants. Even more impressive, it used just 2,000 GPUs in the first phase of training Command A. Businesses that want to use the model on their infrastructure can do so with only two GPUs, a stark contrast to some AI models requiring up to 32 GPUs.
This efficiency-first strategy is what sets Cohere apart. Frosst’s mantra is clear: “ROI, not AGI.” Instead of chasing AGI, which some believe is just around the corner, Cohere is focusing on real-world applications that businesses need today.
Comparing Costs: Cohere vs. DeepSeek
Cohere isn’t the only company demonstrating cost efficiency in AI development. Earlier this year, Chinese AI company DeepSeek shocked the industry by revealing that it built a generative AI model for just $5.6 million using slightly over 2,000 GPUs. This announcement caused panic in the tech sector, leading to a sell-off in AI-related stocks. Many questioned whether AI leaders were overspending on their models.
However, experts believe that the cost estimates from DeepSeek may not be the whole picture. AI model training often involves multiple iterations and costly adjustments before reaching the final version. The $5.6 million cost likely reflects only the last successful training attempt, not the total expenses of development.
Cohere, on the other hand, has been transparent about its costs. The $30 million spent on Command A includes the entire training process, from early experiments to the final working model. This makes it clear that Cohere’s approach to AI development is genuinely more cost-effective.
The Real Question: Will Businesses Use Cohere’s AI?
While building an efficient AI model is impressive, the accurate measure of success lies in adoption. Businesses must see value in using Command A for Cohere to establish themselves as AI leaders.
Cohere is already making moves to expand globally. The company has partnered with Fujitsu to develop a Japanese-language AI model. It has also collaborated with LG CNS, the tech services arm of the South Korean giant LG. These international partnerships highlight the growing demand for AI models that cater to different languages and industries.
Cohere’s focus on business applications gives it an edge in the corporate world. Unlike ChatGPT and other consumer-facing AI tools, which require enormous processing power, Command A is optimized for efficiency. Companies looking for cost-effective AI solutions may find Cohere’s offering highly attractive.
Cohere is proving that building powerful AI doesn’t require billions of dollars. By prioritizing efficiency and real-world applications, the company has created a model that rivals the industry’s most prominent players at a fraction of the cost.
While AI giants continue their costly race toward AGI, Cohere remains focused on what businesses need. With growing global interest and strategic partnerships, the company is positioning itself as a practical, cost-effective AI solutions leader. Whether enterprises will fully embrace Command A remains to be seen, but Cohere’s approach is undoubtedly challenging the status quo of AI development.
Source: Cohere releases low-cost AI model that uses fewer chips
