US-based Ai2 releases new AI model, claims it beats DeepSeek

Another AI company has stepped up to the plate as DeepSeek’s V3 model goes viral, with Ai2 claiming its newest model outperforms its Chinese competitor.

The open-source post-training model, Tülu 3 405B, has been described by the American technology company as being able to surpass the performance of DeepSeek V3.

“The last member of the Tülu 3 family demonstrates that our recipe, which includes Reinforcement Learning from Verifiable Rewards (RVLR) scales to 405B – with performance on par with GPT-4o, and surpassing prior open-weight post-trained models of the same size including Llama 3.1,” Ai2 said on X.

A benchmark was published on the social media site too, with the company looking at their model in comparison to Llama, Nous Hermes, GPT 4o, and DeepSeek.

Benchmarking Tülu 3. Interesting finding: Reinforcement Learning from Verifiable Rewards (RLVR) framework improved the MATH performance more significantly at a larger scale, i.e. 405B compared to 70B and 8B, similar to the findings in the DeepSeek-R1 report. pic.twitter.com/ByCfVlXwiY

— Ai2 (@allen_ai) January 30, 2025

This release follows the launch of Tülu 3 in November, with this new model aiming to demonstrate the scalability and effectiveness of a post-training recipe applied at 405B parameter scale.

Within the announcement, the technology-focused company claims the tool “achieves competitive or superior performance to both Deepseek v3 and GPT-4o, while surpassing prior open-weight post-trained models of the same size including Llama 3.1 405B Instruct and Nous Hermes 3 405B on many standard benchmarks.

“Interestingly, we found that our Reinforcement Learning from Verifiable Rewards (RLVR) framework improved the MATH performance more significantly at a larger scale, i.e., 405B compared to 70B and 8B, similar to the findings in the DeepSeek-R1 report.

“Overall, our results show a consistent edge over DeepSeek V3, especially with the inclusion of safety benchmarks.”

Unlike others, Ai2’s new approach is open source so all components that are necessary to replicate it are freely available and permissively licensed.

A spokesperson for Ai2 was quoted by TechCrunch as saying that the lab believes the model “underscores the U.S.’ potential to lead the global development of best-in-class generative AI models.”

Featured Image: Via Ai2 on X

The post US-based Ai2 releases new AI model, claims it beats DeepSeek appeared first on ReadWrite.