Beam: Reflection's 501B open-weight model for agentic coding
Reflection AI's open-weight sparse MoE (501B total, 23B active) for agentic coding and long-horizon tasks, under Apache 2.0 with a 1M-token context.
Why it matters
A 501B open-weight coding model with permissive Apache 2.0 licensing, hosted for as little as $0.25 per million input tokens, that Reflection says matches Z.ai's GLM-5.2 on its own benchmarks.
What you could build with it
An indie team could self-host Beam on rented GPUs to run a long-horizon coding agent overnight, keeping the weights and the work in-house instead of paying per-token frontier prices.
Does it hold up?
Usable now as a cheap hosted open-weight coding model, but expect rough edges: reasoning cannot be disabled, the API had two full outages in October, and safety tests show gaps under sustained pressure.
Built with Beam
- Reflection AI Beam Pricing: $1.32 per 1,000 Answersmxchat.ai · Independent cost test across five OpenRouter hosts: $1.32 per 1,000 support-bot answers on Nebius, with reasoning that cannot be switched off.
- Reflection publishes Beam safety research showing gaps under persistent attackssuperpowerdaily · Beam refused 89% of harmful requests in Reflection's Iris tests but complied 15% of the time under sustained multi-turn adversarial pressure.
- Best Open-Weight Models in 2026: Ranked by Job and Licencehokai.io · Beam picked as the big Apache 2.0 model with no licensing strings, scoring 18 on the Artificial Analysis Intelligence Index.
Learn more
First spotted on article: source.
More AI models
VoiceStudio: open-source, fully-local ElevenLabs alternative for voice cloning, dubbing, and audiobooksdebpalash/VoiceStudio is a desktop app that runs voice cloning, voice design, video dubbing, dictation, transcription…models · JEV 0.75Soup: fine-tune an 8B LLM on a 4 GB laptop GPU from one YAML fileAn open-source CLI that fine-tunes LLMs from a single YAML file, with layer streaming that trains an 8B model on a 4…models · JEV 0.73Tencent Hunyuan Hy3: open 295B MoE with 21B active under Apache 2.0Tencent open-sourced Hy3, a 295B-parameter mixture-of-experts model with only 21B parameters active per token, under…models · JEV 0.73NVIDIA Isaac Lab 3.0 GA + Isaac OS open runtime + Agile ONE SN humanoidMajor Isaac platform expansion at GTC Berlin Oct 21: Isaac Lab 3.0 is generally available as the open-source robot…models · JEV 0.73NVIDIA Nemotron 3.5 Retriever collection: Embed 8B/1B, Rerank 8B, Parse 631MNVIDIA's Oct 21 collection drop for agentic retrieval: Nemotron-3.5-Embed in 8B BF16, 1B BF16 and 1B NVFP4…models · JEV 0.72Kimi K3.1 Pro: Moonshot's 2.8T open-weight flagship officially launchesMoonshot AI officially launched Kimi K3.1 Pro on Oct 23, a mixture-of-experts model with 2.8 trillion total / 104…models · JEV 0.72
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.