What is mattwood.fyi?

A live, structured feed of riffs and links from Matt Wood, Chief AI & Technology Officer at AWS — readable by people and queryable by agents as a typed knowledge graph.

Is there an API and is it free?

Yes. A public, read-only query API is available with no authentication and no API keys. See https://mattwood.fyi/openapi.json and https://mattwood.fyi/docs.

How should an AI agent use this site?

Fetch https://mattwood.fyi/feed.json for the full structured feed, or query the API for keyword/semantic search and the connection graph. See https://mattwood.fyi/agents/ for when-to-use guidance.

Introducing Beam: Reflection's 501B Open-Weight Model

Reflection releases Beam, a 501 billion parameter sparse Mixture-of-Experts model optimized for coding, reasoning, and agentic tasks, trained with large-scale reinforcement learning and achieving competitive performance with strong inference efficiency compared to similar open-weight models.

reflection.ai →

Connections

Related to

Atria Dawn: Agentic Superintelligence Foundation Model Both Beam and Atria Dawn are positioned as frontier foundation models targeting agentic superintelligence/agentic tasks, competing in the same open-weight agentic model space
MiMo-V2.6 | Xiaomi Both are large open-weight models with strong reasoning capabilities released by non-OpenAI players, representing the competitive landscape of frontier open-weight reasoning models

Supports

The Ultimate Guide to Multi-Harness RL Beam's training methodology relies heavily on large-scale reinforcement learning, directly illustrating the multi-harness RL techniques described in that guide
Inside the Inference Hardware Revolution Of 2026 A 501B sparse MoE model like Beam exemplifies the inference hardware revolution of 2026, requiring specialized distributed inference infrastructure discussed in that piece
SWE-bench Science: Coding Agents in Scientific Software Engineering Beam's optimization for coding tasks makes it directly relevant to SWE-bench Science benchmarks evaluating coding agents in scientific software engineering
Thinking Before Thinking: Scaling Agentic Inference Through Meta-Reasoning Beam's design for agentic tasks with RL-trained reasoning supports the thesis about scaling agentic inference through meta-reasoning, as large RL-trained MoE models enable the kind of pre-inference planning described

Mistral Large 4 Reflection's 501B open-weight model and Mistral Large 4 represent competing approaches to frontier model development in 2026, with different strategies around openness and scale

Type:	link
First published:	Monday, 5 October 2026 at 23:21 UTC
Location:	https://reflection.ai/blog/introducing-beam
URL:	https://mattwood.fyi/i/835b1d39

I'm Matt Wood, and this is For Your Information. A live list of riffs and links for you and your agent, drawn from what I'm reading, noticing, questioning, concluding, and revising.

Connections