Reflection launches Beam: 501B open-weight MoE trained on 23.8T tokens with 100M+ RL rollouts
gethackteam · x · 2026-10-06
Reflection announced Beam, its first open-weight model: a sparse MoE with 501B total / 23B active parameters, built for coding, reasoning, and agentic workloads.
Training: pretrained on 23.8T curated tokens (web + licensed proprietary data), matching or beating similar-sized open base models; a high-compute RL run generated 100M+ rollouts on 10.5K NVIDIA GB300 GPUs over 4 weeks.
Positioning: advances the Western open-weight frontier on coding and agentic tasks — competitive with GLM 5.2 and approaching Qwen 3.8-Max, though frontier open models like Kimi K3 stay ahead on raw capability; Beam's edge is inference-time efficiency. Benchmark charts cover agentic coding/terminal, reasoning, and tool calling.
Rollout: final red-teaming underway, early access open now; weights, tech report, model card, and developer artifacts ship later this month.
Related event: Reflection AI Debuts Beam, a 501B-Parameter Open-Weight Model(20 posts)→
More from Infra
- SpaceX files for 32.4-mile Florida gas pipeline to make Starship methane on-site — DimaZeniuk · 2026-10-06
- Inference Companies Ship Gateways, Gateway Companies Ship Inference — the Lines Are Blurring — michellechen · 2026-10-06
- Can an 8GB Radeon 5700 Run Image-to-Video? ComfyUI Low-VRAM Question — Fluffy-Composer9675 · 2026-10-06
- Pairing a Blackwell RTX 4000 24GB with an Old RTX 3060 12GB for Local LLMs — Otherwise-Tangelo-52 · 2026-10-06
- Local LLM for VFX: 2x DGX Spark vs M5 Ultra 256GB for Blender/Houdini via MCP — abrasmel · 2026-10-06
- Modal Runtime keynote ships Endpoint Candidates, Clusters and VM Sandboxes — graceisford · 2026-10-06