All tasks / Models

Beam 1.0 (501B-A23B)

Reflection AI's first open-weights MoE LLM (Apache-2.0) with 501B total and 23B active parameters.

Reflection AI
Open weights
Released 7 Oct 2026

Where it ranks

C
Chat & assistants
Models · Local · 30 on Artificial Analysis Intelligence Index (secondary benchmark)
C
Coding
Models · Local · 1,437 on LMArena WebDev / Code Arena (overall)
C
Workflow automation
Models · Local · 48.5% on AutomationBench-AA (Artificial Analysis)

Facts

Vendor: Reflection AI
Family: Beam
Released: 7 Oct 2026
Weights: Open: download and run it yourself
License: Apache-2.0
Parameters: 501B (23B active)
GPU memory to run locally: ~287 GB (4-bit ≈ 501 B × 0.57 GB + 1.5 GB ≈ 287 GB, estimate.)
API price: $1 in / $3 out per 1M tokens (OpenRouter listed price)
Access: Free (open weights)
Last verified: 19 Oct 2026

Price from openrouter.ai, 19 Oct 2026.

Benchmark results

Benchmark	Result	Entry on the leaderboard	As of	Source
Artificial Analysis Intelligence Index	30	`Beam 1.0` Raw value from the page's embedded model data; the site displays it rounded to an integer.	19 Oct 2026	artificialanalysis.ai
AutomationBench-AA (Artificial Analysis)	48.5%	`Beam 1.0` AutomationBench-AA Score (share of each task's objectives completed, a task with any guardrail violation scoring zero); raw value from the page's embedded model data (displayed rounded); rank 27 of 32; Tasks Completed 17.05%, Objectives Completed 77.92%; no last-updated date shown, so as_of = retrieval date.	19 Oct 2026	artificialanalysis.ai
LMArena WebDev / Code Arena (overall)	1,437	`beam-1.0-501b-a23b` From LMArena's official leaderboard-dataset (webdev = Code Arena, split latest, category overall, published 2026-10-18), which mirrors https://arena.ai/leaderboard/webdev. Rank 79; 2907 votes; 95% CI 1424.0-1449.6.	18 Oct 2026	huggingface.co

Where you can use it

No tracked tool lists this model yet; as open weights it runs in local runtimes.

Sources