Laying out the streets…STEP 1 OF 3

Beam · Office #44,663

Highly efficient agentic 501B parameter open model

In the Run AI locally neighborhood (AI and agents). In town since 19 Oct 2026.

Nobody has claimed this office yet. Tablif records every new product and gives it an office; when its owner claims it, the lights go on.

Is this yours? Claim it

What Tablif read on Beam's site

Beam describes itself as: “Introducing Beam: Reflection’s 501B open-weight model — Reflection”

What it is

Calls itself “model”.

  • Calls itself: model
“Beam is a sparse Mixture-of-Experts model with 501 billion total parameters, 23 billion active, built for coding, reasoning, and agentic workloads.”

Where it puts itself

Claims “first”. Rules out “text-only”.

  • Superlative claim: first
  • Says it will not: text-only
“Where frontier open models like Kimi K3 remain ahead on raw capability, Beam's advantage is efficiency at inference time.”
“We are introducing Beam, Reflection’s first open-weight model.”

What it costs

Sales on request.

  • Contact sales: yes
“Get in touch”

Who it sells to

Sells to businesses. Built for enterprise.

  • Sells to: businesses
  • Company size: enterprise
“…making Beam a powerful workhorse model for enterprise coding and agentic workloads.”

How it runs

Distributed through the Hugging Face. Offers an API. Open source. AI is the core of the product. Built with Google Analytics.

  • Distributed via: Hugging Face
  • API: yes
  • Open source: yes
  • AI role: core to the product
  • Analytics: Google Analytics
“Weights Pull the weights from Hugging Face (opens in a new tab)”
“Get an API key and start building with Beam on Reflection's API”

What it works with

Names GLM 5.2, Gemma-4 and 4 more on its pages.

  • Names: GLM 5.2 · Gemma-4 · Kimi K3 · +3
“Beam is a sparse Mixture-of-Experts model with 501 billion total parameters, 23 billion active, built for coding, reasoning, and agentic workloads.”
“Beam advances the Western open-weight frontier and is competitive with larger open models like GLM 5.2 and approaching Qwen 3.8-Max on coding and agentic tasks.”

What it offers as proof

Live, no waitlist. Claims “501 billion total parameters”. Instant signup. Publishes a blog and a careers page.

  • Number claimed: 501 billion total parameters
  • Instant signup: yes
“You can sign up here for early access to the model.”
“Beam is a sparse Mixture-of-Experts model with 501 billion total parameters, 23 billion active, built for coding, reasoning, and agentic workloads.”

Sources

Tablif read these pages, last on 19 Oct 2026. Everything above is what the product says about itself; Tablif does not verify every claim.

Similar products in Tablif Town

Explore Tablif Town