Reflection AI says Beam is built for businesses and developers who need help with coding and AI agent tasks, and it aims to be cheaper than a lot of competing options. The startup contends Beam can hold its own against leading models in the US and China, and it plans to share more technical details later this month.
According to the company, Beam posts comparable scores to GLM-5.2 on advanced reasoning benchmarks, referencing the flagship open model from China's Z.AI. This rollout follows Reflection's exit from stealth in the prior year, when it set its sights on open AI models. Investors include Nvidia Corp. and others.
Open-weight models change who can build on frontier AI and at what cost. Market Briefs covers the AI landscape free every weekday.
Chinese players such as DeepSeek and Moonshot AI have been closing the gap by shipping cheaper, customizable open models, even as China is still broadly seen as behind the US in AI. By contrast, top US outfits like OpenAI and Anthropic PBC have mostly concentrated on closed or proprietary models as they build out enterprise offerings.
On the funding and infrastructure front, the Wall Street Journal previously reported that, according to earlier reporting in the Wall Street Journal, Reflection discussed a $2.5 billion raise valuing the company at $25 billion. The company inked agreements worth billions with Elon Musk's SpaceX and with cloud firm Nebius Group NV, locking in the compute needed to train and run its models.
Beam's debut could also align with Washington's push to expand AI from the US and its allies abroad. Earlier this year, Reflection took part in a US-facilitated pact aimed at establishing sovereign AI in Korea. Government officials are pursuing further deals to boost US tech exports and curb the diffusion of Chinese options.
If open-weight models keep improving while staying cheaper, more companies may favor customizable setups over closed, one-size-fits-all tools. That can influence where budgets go for compute, cloud credits, and software. Watch which vendors back open models and how governments stitch together cross-border AI deals, because those choices often show up later in enterprise spending.
Each open release shifts the competitive balance a little further. Join Market Briefs free and track the releases.
