signalTechCrunch AI2026-10-06
Reflection debuts Beam, an open-weight AI model to rival Chinese models at lower compute cost
Reflection AI launched Beam, a 501-billion-parameter open-weight model with 23 billion active parameters, pretrained on 23.8 trillion tokens and a 1 million token context window. The startup claims Beam matches Z.ai's GLM-5.2 on reasoning benchmarks while using 3-4x less inference compute, though results are unverified. Reflection has raised $4.7 billion, with a $25 billion pre-money valuation, and secured over $7 billion in compute deals with SpaceX and Nebius for Nvidia GB300 chips through 2029.
- for who
- Enterprises, public sector, developers, and sovereign nations seeking cost-effective open-weight AI models.
- why now
- Beam's open-weight release this month challenges cheaper Chinese models, pressuring Western AI rivals now.
- what changes
- These groups can adopt a Western open-weight model that rivals Chinese models at lower inference cost, enabling custom local AI factories without relying on closed labs.
- to do
- Watch for Beam's weight release this month and evaluate it against GLM-5.2 and Inkling for reasoning and coding tasks.
key points
- Beam is 501B total parameters, 23B active, pretrained on 23.8T tokens
- Claims 3-4x less inference compute than rivals like GLM-5.2
#open-weight model#AI competition#inference cost
score
score 7 out of 10. 0-10: how dense the facts are, multiplied by how much you can do with them after reading. 8+ means the topic's evidence bar is met: benchmarks and availability for a new model, amount and investors for a funding round, revenue figures for a solo-money story. Below 5 an item does not enter the digest. A press release scores 3 or less, a reprint loses 2, anything older than 14 days loses 1, a headline that misleads loses 3.
read the source