AI hardware and software vendor Cerebras Systems launched Alibaba's QWEN3-235B inference model adapted to Cerebras' inference platform on Tuesday.
The vendors have also revealed partnerships with other vendors, such as concepts and data robots.
At the Raise Summit Conference in Paris, Cerebras introduced the inference model of China's Tech Giant with a 131k context window in the inference platform Cerebras Inference Cloud.
Alibaba and other partnerships
According to AI vendors, its wafer-scale engine speeds up the QWEN3-235B, reducing response time from 1 to 2 minutes to 1 to 2 seconds. Celebras also extended the QWEN3-235B context window from 32K to 131K tokens, allowing the model to process dozens of files and tens of thousands of code at the same time.
Cerebras offers the model with $0.60 input tokens per million and $1.20 output tokens per million. It states that this is less than a tenth of the equivalent closed source model. Openai's O3 inference model is $1 million in input tokens, $8 per million.
In addition to this offering, Cerebras has introduced several partnerships, including Docker, a vendor of open source platforms that helps developers build and run applications. With Docker Compose and Cerebras, developers can deploy a multi-agent AI stack within seconds, Cerebras said.
Cerebras also revealed that SYFTR, Datarobot's new open source AI/machine learning framework for automating agent workflows, is currently using Cerebras inference. The integration enables Datarobot customers to build high quality, low latency agent applications. Celebras reasoning can also embrace Face's Smolagents Library, allowing developers to create agents, use tools and execute code for that reason. The concept, which is a Workspace platform, now uses Cerebras' AI inference technology to provide AI, conceptual AI to work.
Extend its existence
Through the partnership, Cerebras is expanding its scope as they are trying to compete with vendors such as GROQ, AMD and Sambanova.
“The cerebrum seems to be doing a good job of expanding ecosystems,” says Karl Freund, founder of the Cambrian AI. “[It] Make your products more accessible and reliable in the market. ”
He found Celebras, a vendor that started on a small scale in 2015 and now offers ultra-fast wafer-scale engine systems for training generation AI models of between $20,000,000,000,000,000,000,000,000,000,000,000, to be difficult to accept in the market as it lacks much software that comes with hardware.
“They are known as the fastest AI machines out there, but they are also known as the most expensive machine systems,” Freund said. He added that Celebras needs to demonstrate its price performance, and that the partnership will help.
According to Addison Snell, CEO of Intersect 360 Research, vendors also face the challenge of gaining more awareness and consideration for processors compared to the dominant AI chip maker Nvidia.
“It's a very high sales cost and it's establishing that level of interest. There are a lot of costs,” Snell said. Additionally, Nvidia provides hardware with software components. This is an approach that many people like Celebras are trying to pursue and keep up, Snell said.
QWEN3-235B
Celebras stands out because of the expanded context window when it comes to QWEN3-235B deployment.
“It takes it from the toys to the actual enterprise platform,” Freund said. He added that a larger context window would be useful for people who want to write code to increase productivity.
“If we can run a big context window that is important for coding and agent AI, then if we can do that for a tenth of the price, I think there's something that makes a difference,” Freund continued.
The QWEN3-235B deployment shows how Celebras thinks about the international market.
“Top AI process providers are looking at the global market for processors and how they compete internationally,” Snell said. “In the long run, national sovereignty efforts, particularly in government-funded or managed AI data centres, may make it even more difficult, but for the time being, it seems like a fair play.”
Meanwhile, Openai, Microsoft, and Anthropic have collaborated with the American Federation of Teachers to establish the National Academy of AI Leadership. The new $23 million initiative aims to train 400,000 educators over the next five years. The Academy includes practical workshops for educators. Its headquarters is in Manhattan.
Microsoft also introduced deep research on Azure AI Foundry on July 7th. This API and SDK allows developers to build agents that can plan, analyze and synthesize information across the web.
Esther Shittu is an Informa TechTarget News Writer and Podcast host that covers artificial intelligence software and systems.
