Nvidia-backed Reflection AI launches Beam, says it needs 3-4 times less compute

Nvidia-backed Reflection AI challenges Chinese dominance in open-weight models

Reflection AI has released Beam, an open-weight artificial intelligence model that it says is designed to use substantially less computing power while reasoning through problems.

The Nvidia-backed US start-up said its first model is aimed at coding, reasoning and agentic tasks.

Focus on efficiency and open access

Reflection said Beam requires three to four times less inference compute than comparable open models when reasoning through problems.

Early third-party testing suggests that the system could be among the most token-efficient open models in its class.

The release is positioned as a challenge to China’s dominance in freely available artificial intelligence models.

How Beam compares with other models

Reflection AI debuted Beam on Monday and described it as “competitive” with GLM-5.2 from Z.ai, also known as Zhipu AI.

The company said Beam was “approaching” Alibaba Group Holding’s Qwen3.8-Max on coding and agentic workflows.

However, Reflection acknowledged that Moonshot AI’s Kimi K3 remained ahead in “raw capability”.

Alibaba owns the South China Morning Post.

Source: South China Morning Post