Home · Technology · Oct 5 archive
Reflection Launches Beam AI Model to Rival Chinese Competitors
Confirmed
In Short: Reflection, a startup based in Brooklyn, has officially launched Beam, its first frontier, open-weight AI model, positioning it against both closed labs like Anthropic and OpenAI and popular open models from Chinese developers.
Reflection, a startup based in Brooklyn, has officially launched Beam, its first frontier, open-weight AI model, positioning it against both closed labs like Anthropic and OpenAI and popular open models from Chinese developers.
This summer, Reflection secured deals worth more than $7 billion with SpaceX and Nebius, ensuring access to Nvidia’s GB300 chips through 2029.
According to the company, Beam matches the performance of leading Chinese open models on advanced reasoning benchmarks but at a significantly lower cost.
In a blog post, Reflection described Beam as a text-only mixture-of-experts model trained on high-compute reinforcement learning, making it effective at reasoning, coding, and agentic tasks.
Beam is a 501-billion-parameter model with 23 billion active parameters, and it claims to outperform today’s leading Western open models while using 3-4 times less inference compute.
The company’s performance claims have not been independently verified, but they suggest Beam could be a strong contender in the race to build a Western alternative to Chinese models like DeepSeek, Qwen, and Z.ai.
Beam is aimed at enterprises, sovereign nations, and developers, positioning itself as a versatile 'workhorse model' for various sectors.
What this adds
The launch of Beam by Reflection could intensify competition in the AI model market, particularly between Western and Chinese developers.
What's confirmed
- Reflection, a startup based in Brooklyn, has officially launched Beam, its first frontier, open-weight AI model, positioning it against both closed labs like Anthropic and OpenAI and popular open models from Chinese developers.
- This summer, Reflection secured deals worth more than $7 billion with SpaceX and Nebius, ensuring access to Nvidia’s GB300 chips through 2029.
- According to the company, Beam matches the performance of leading Chinese open models on advanced reasoning benchmarks but at a significantly lower cost.
- In a blog post, Reflection described Beam as a text-only mixture-of-experts model trained on high-compute reinforcement learning, making it effective at reasoning, coding, and agentic tasks.
- Beam is a 501-billion-parameter model with 23 billion active parameters, and it claims to outperform today’s leading Western open models while using 3-4 times less inference compute.
- The company’s performance claims have not been independently verified, but they suggest Beam could be a strong contender in the race to build a Western alternative to Chinese models like DeepSeek, Qwen, and Z.ai.
- Beam is aimed at enterprises, sovereign nations, and developers, positioning itself as a versatile 'workhorse model' for various sectors.
What's still developing
- Reflection’s announcement confirms reporting from Axios over the weekend that the startup was close to a launch.
- But for us audiences from West Asia and North Africa (WANA), regions that have lived through the wars depicted in the film, the story it tells provokes numerous reflections.
- But it fails to tell a Chinese story from a full WANA perspective.
- Chinese cinema must tell that story in a louder voice.