Reflection AI launches open-weight coding model Beam
The Nvidia-backed startup says Beam matches leading US and Chinese AI models on coding and agent tasks at lower cost.
Published
Reflection AI has released Beam, an open-weight AI model built for coding and AI-agent tasks. The company says it can match leading US and Chinese models while costing businesses and developers less to run.
Reflection AI is backed by Nvidia. Beam runs on Nvidia infrastructure.
The release speaks to competition at the software layer rather than to chip demand itself. Investors have pointed to other factors as bigger near-term swings for Nvidia: Jensen Huang's guidance for doubled shipments and the unresolved question over China export rules. Both carry more weight than any single open-weight model launch. Cheaper, open alternatives like Beam could eventually pressure what customers pay for proprietary AI software, but that would be a second-order effect on the chips underneath it, not a change to compute demand now.
Nvidia shares hit their first record high since May on Friday, rising 2.5%. Morgan Stanley named Nvidia its top semiconductor pick the same day. On Thursday, Nvidia and SoftBank each completed $10bln investments in OpenAI, bringing each firm's total pledge to $30bln.