Reflection AI says its Beam model was trained through two large computing runs: pretraining on 6,144 NVIDIA GB300 NVL72 GPUs, followed by four weeks of reinforcement learning on 10,500 GB300 GPUs. The company also reports using about 1.3 billion sandboxes for training and grading during that four-week reinforcement-learning run—roughly 46.4 million a day when the approximate total is divided by 28. Beam is a 501-billion-parameter sparse model, but Reflection’s claim that it uses less inference compute than some rivals is an estimate, not a measured end-to-end cost comparison.
What Beam is—and what the training figures mean
Reflection AI introduced Beam on October 5, 2026, describing it as its first open-weight model. It is a sparse Mixture-of-Experts (MoE) model with 501 billion total parameters and 23 billion active parameters, designed for coding, reasoning, and agentic workloads. In an MoE model, only a portion of the model’s parameters are active for a given token; the total parameter count and active parameter count therefore describe different things.
As an Amazon Associate I earn from qualifying purchases.
The headline-scale GPU figures refer to separate stages, not one simultaneous cluster: Reflection says pretraining used 6,144 GB300 NVL72 GPUs, while reinforcement learning (RL) ran on 10,500 GB300 GPUs for four weeks. The company reports the RL phase generated more than 100 million rollouts and used approximately 1.3 billion sandboxes for training and grading. Reflection has not provided independent confirmation of these training totals; they are its account of its own work.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallHow Reflection says it trained Beam
Pretraining: data, compute, and recovery
Reflection says Beam’s base model was pretrained on 23.8 trillion curated tokens from web sources and proprietary licensed datasets. Its described data pipeline used quality classifiers for web, code, and STEM material; fine-grained quality tiers; language-specific code filters; and processing for technical PDFs. The company says parsing, deduplication, and curation removed about 95% of raw internet tokens while retaining roughly 1.8 trillion high-quality tokens that conventional techniques would have missed. Those dataset figures and characterizations come from Reflection and are not independently verified.
#1 Best Overall
- Professional AI & Creator Workstation: AMD Radeon AI PRO R9700 GPU with 32GB GDDR6 is engineered for AI development, professional content creation, and compute-intensive workloads.
- Massive 32GB Memory Capacity: 32GB of GDDR6 memory on a 256-bit bus provides ample bandwidth for large AI models, 8K video editing, and complex 3D rendering.
- Advanced RDNA 4 with AI Accelerators: 64 Compute Units with 3rd Gen Ray Tracing and dedicated 2nd Gen AI Accelerators for groundbreaking AI performance and visual computing.
- Professional Blower Cooling: Efficient single blower design exhausts heat directly out of the chassis, ideal for multi-GPU workstation and server configurations.
- Enterprise-Grade Thermal Solution: Vapor chamber heatsink with industrial Honeywell PTM7950 thermal interface material ensures reliable cooling under sustained professional loads.
The company reports that end-to-end pretraining finished in under four weeks on a cluster of 6,144 NVIDIA GB300 NVL72 GPUs. It describes using in-house scheduling, node-health monitoring, corruption detection, and semi-automatic rewind systems. Reflection reports nine semi-automatic rewinds and 92.3% goodput toward the end of the run; its announcement does not independently validate those operational figures.
Reinforcement learning: rollouts and environments
For RL, Reflection reports using 10,500 NVIDIA GB300 GPUs over four weeks, with a maximum context length of 256K tokens. The run produced more than 100 million rollouts—model-generated attempts or trajectories used in training—and drew on one million coding, agentic, and STEM environments. Reflection says those environments were sourced primarily through synthetic-data pipelines, supplemented by vendor and open-source sources.
Rank #2
- 【AI Max+ 395 AI Workstation】16 cores, 32 threads, up to 5.1 GHz boost and 80 MB cache. Integrated Radeon 8060S graphics with 40 CUs, RDNA 3.5, delivers performance close to RTX 4060/4070 laptop GPUs. Triple-engine design(CPU+GPU+XDNA 2 NPU) with up to 126 TOPS total, including 50+ TOPS dedicated NPU for local AI inference and machine learning acceleration. Ideal for AI development, content creation, virtualization, data analysis, and demanding multitasking. Compact, high-performance workstation.
- 【256-bit LPDDR5X MAX 128GB】The LPDDR5X onboard memory reaches 8400 MT/s - 1.5x faster than DDR5 SODIMM. Unlock the full potential of your graphics with massive 128GB memory pooling. This system allows you to manually assign up to 128GB of the onboard RAM to serve as video memory (VRAM) directly within the BIOS setup, delivering unparalleled performance for 4K video editing, and AI model training without the need for a discrete graphics card.
- 【Lastest GPU 8060S & XDNA 2 NPU】Built on the RDNA 3.5 architecture, the AMD Radeon 8060S Graphics iGPU features 40 compute units (2,560 stream processors). It delivers performance on par with NVIDIA's mobile RTX 4070, efficient encoding/decoding for AVC, HEVC, VP9, and AV1 video codecs. And It can connect 4 screens via HDMI & DisplayPort & Full Featured USB4 x2 to efficiently handle your tasks and meet your specific needs. Supports 8K/4K resolution displays.
- 【Dual LAN (2.5GbE+10GbE)& WiFi 7】The computer has double LAN, one is 2.5GbE (I226), the other is 10GbE(AQC113). provides more applications, such as firewall, soft routing, multichannel aggregation. Built-in WiFi module, support WiFi 7 and Bluetooth5.4. Known as 802.11be, Wi-Fi 7 promises up to 46Gbps theoretical throughput, making it 4.8x faster than Wi-Fi 6. and computer has 4 built-in NVMe SSD slots, 1 SD card slot, allowing you to expand its storage capacity.
- 【Engineered to Endure】The computer measures 7.13 x 7.24 x 2.99 inches. AI mini pc is encased in a premium all-aluminium chassis. Dual turbo CPU fans deliver silent, ultra-efficient cooling, To enable the computer to maintain stable operation for a long time. We offer up to 2 years warranty and lifetime professional customer service. Please feel free to contact us if any issues happened. thanks
The company describes using asynchronous policy gradients and methods for learning from rollouts generated more than a day earlier while managing policy staleness and numerical differences between training and inference. It reports an average of 110,000 concurrent rollouts and support for as many as 170,000 concurrent sandboxes. These concurrency figures are separate measures: one counts rollouts, the other sandbox capacity.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minuteWhat “46.4 million sandboxes per day” represents
Reflection’s announcement gives an approximate total of 1.3 billion sandboxes for training and grading across the four-week RL run; it does not report a daily sandbox count. Dividing that approximate total by 28 days gives about 46.4 million per day. This is a rough average derived from the company’s figures, not evidence that the same number of sandboxes was created or used each day.
Rank #3
- [ Maximum AI Compute Power ] Dominate complex workloads with the ASUS ESC8000A-E13. This 4U rack server is a powerhouse engineered for mass-scale AI, machine learning, and deep training. Featuring support for dual AMD EPYC 9005/9004 processors and up to eight dual-slot GPUs, it delivers the raw computational muscle required to train LLMs and run complex simulations effortlessly. Accelerate your data science pipeline and transform raw data into actionable intelligence faster than ever.
- [ Advanced Thermal Efficiency ] High performance demands elite cooling. The ESC8000A-E13 features a cutting-edge aerodynamic design with independent CPU and GPU airflow tunnels. Equipped with redundant hot-swap fans and optimized for liquid cooling integrations, this 4U server ensures maximum uptime under heavy, sustained workloads. Keep your data center running cool, quiet, and highly efficient while preventing thermal throttling during mission-critical enterprise operations.
- [ Scale with Flexible Storage ] Future-proof your infrastructure with unmatched storage and expansion flexibility. This offers comprehensive front-panel drive bays supporting Gen5 NVMe, SAS, or SATA drives alongside multiple PCIe 5.0 slots. Designed as a high-density 4U server capable of housing eight dual-slot GPUs: NVD H200, RTX PRO 6000 Blackwell, RTX PRO 4500 Blackwell or AMD Instinct MI350P PCIe Card, each supporting up to 600 watts.
- [ Enterprise-Grade Reliability ] Minimize downtime and secure your ecosystem with server-grade redundancy. The ESC8000A-E13 is built for 24/7 continuous operation, boasting 2+2 redundant (3200W total) 80 PLUS Titanium power supplies and integrated ASUS ASMB11-iKVM for comprehensive out-of-band management. Ideal for cloud service providers, rendering farms, and large enterprise infrastructure, it combines robust physical hardware with smart remote monitoring to safeguard your digital assets.
- [Reliability Guaranteed] Shop with total peace of mind knowing that every new computer component we sell is backed by our EPC 3-year warranty. Whether you are investing in high-speed DDR5 RAM or a powerhouse GPU, we protect your build against defects and performance failures. We stand firmly behind the quality of our hardware, ensuring that your setup remains fast, stable, and secure for years to come.
What Reflection reports about the infrastructure
Reflection says its platform processed more than one billion sandbox creation requests across over 20 clusters, two clouds, and four regions. It reports that 90% of new sandboxes were ready in under 10 seconds and that dynamic packing kept batches 99.99% full on average. These are platform-wide company figures associated with the training effort, not third-party measurements.
For moving model weights to the inference fleet, Reflection reports a median delivery time of about 12 seconds. It says hierarchical transfer over RoCE and NVLink reduced cross-rack traffic by 75% and made fleet-wide adoption 2.2 times faster than direct pulls by every replica. The company also says it handled 71 inference incidents without terminating the training job, with median inference-capacity recovery of eight minutes and lost capacity equal to 0.02% of elapsed serving GPU-minutes. Each of these operational results is reported by Reflection.
Rank #4
- AMD socket sTR5 supports up to 96-core CPUs: Ready for AMD Ryzen Threadripper PRO 7000 WX-Series Processors.
- Ultrafast connectivity:Seven PCIe 5.0 x16 slots, dual 10 Gb LAN ports, four M.2 slots, two rear USB4 40Gbps Type-C and SlimSAS NVMe support.
- CPU and memory overclocking: Support for up to 2TB ECC R-DIMM DDR5 memory modules (1DPC)
- Robust power and thermal design: 32 power stages with two 8-pin power connectors for the CPU, massive VRM cooling, chipset and M.2 heatsinks with active fans, and M.2 thermal pad.
- PCIe Q-release Slim: Remove the graphics card by directly pulling it up, instead of pressing a PCIe latch.
What the benchmark and efficiency claims do—and do not—show
Reflection reports Beam scores of 80.9 on SWE-bench Verified, 80.1 on Terminal-Bench 2.1, 97.8 on AIME 2026, and 90.5 on GPQA Diamond. These results span software engineering, terminal tasks, mathematics, and scientific reasoning; a score on one benchmark is not directly comparable to a score on another. The announcement includes other coding, reasoning, tool-use, search, and general-capability benchmarks, and marks some comparison entries as not reported. Its scores and benchmark setup are company-reported rather than independently validated in the announcement.
Reflection says Beam scored comparably to GLM 5.2 on advanced reasoning benchmarks while using an estimated three to four times less inference compute. It says its comparison with models in the 2T-plus parameter family, including Qwen 3.8-Max, showed larger efficiency differences. These claims should be read as Reflection’s estimates, not as proof Beam is universally cheaper, faster, or better in deployment.
- Task capability: Compare results on the same benchmark and version, with attention to the evaluation setup. Different benchmarks test different abilities.
- Compute estimate: Reflection estimates generation FLOPs using active parameter count and mean generated tokens. It says the estimate excludes prompt prefill, context-dependent attention, and serving overhead, so it is not an end-to-end cost measurement.
- Reasoning effort and token use: Reflection says users can set reasoning effort, trading shorter responses against more reasoning on demanding tasks. Actual token consumption and deployment cost will depend on usage.
- Evidence quality: The announcement provides company-reported benchmark results and estimates, not independent replications or measured costs from multiple deployments.
Safety work and the announced release plan
Reflection says it trained a separate safety and alignment model using its own supervised fine-tuning and RL pipeline, then combined teacher capabilities through multi-teacher on-policy distillation. It describes adversarially generated training prompts and safety scenarios covering single-turn, multi-turn, jailbreak, and agentic settings. The October 5 announcement did not include safety evaluation results; Reflection said those results and internal evaluation tools would be published with a technical report.
At announcement time, Reflection said Beam was undergoing final red-teaming and evaluations, with early access offered to a select group. It planned an October 2026 release of the weights under Apache 2.0, alongside documentation and developer tools for running, evaluating, and fine-tuning the model. That was a plan announced on October 5, not confirmation that the weights had subsequently been released or that the announced terms and materials were available.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
