# Baseten — Beam: Efficient reasoning for coding and agentic workloads

- Company: Baseten (baseten.co)
- Announced: 2026-10-16T13:51:20+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://www.baseten.co/blog/reflection-beam-launch
- Record: https://forck.live/items/23542-beam-efficient-reasoning-for-coding-and-agentic-workloads
- Subject: Baseten
- Open weights: yes
- Models affected: Beam
- Context window: 1 million tokens

Baseten announced the availability of Beam, Reflection's first open-weight model, on its platform. Beam is a 501B-parameter mixture-of-experts model with 23B active parameters per token and a 1M-token context window, designed for efficient reasoning in coding and agentic workflows. It is available through Baseten Model APIs and Dedicated Inference.

## Evidence

Verbatim from https://www.baseten.co/blog/reflection-beam-launch:

> Beam is an autoregressive transformer model that accepts text inputs and generates text outputs (mainly in English). It has a mixture-of-experts (MoE) architecture with 501B total parameters, 23B active parameters per token, and a context window of up to 1 million tokens.

---

Record: https://forck.live/items/23542-beam-efficient-reasoning-for-coding-and-agentic-workloads
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md