Beam

501B OPEN APACHE

Reflection's first open-weight model: 501B/23B-active MoE for coding, reasoning, and long-horizon agentic work, Apache 2.0 with BF16/FP8/NVFP4 weights plus a speculative-decoding drafter; token-efficient (~3x fewer reasoning tokens for comparable quality), SWE-bench Verified 81.6%, Terminal-Bench v2.1 80.1%, SWE-Bench Pro v2-Hard 76.5%; ~1.3B RL sandboxes across training; $0.25/$1.00 per 1M tokens.

Context

1,000,000 tokens

Architecture

501.3B Total / 23.4B Active MoE (256 experts)

Official API (1M)

$0.25 in / $1 out (Open)+ Weights Free to Self-Host

License

Apache 2.0

Published benchmarks · lab-reported

Vendor-published figures, not independently reproduced. Sources in Verification below.

SWE-BENCH

81.6% (Verified)

TERMINAL-BENCH

80.1% (Terminal-Bench v2.1)

Quickstart

curl -s https://modelregistry.tirup.in/api/cli?model=beam-1

OFFICIAL ANNOUNCEMENT WEIGHTS / REPOSITORY API DOCS

BACK TO LEDGER

Sources

Beam is now available · Reflection
Beam model card · Reflection
Beam open weights (Hugging Face collection) · Reflection
Beam listing and pricing · OpenRouter

Report an error in this record