Reflection has released Beam, an open-weight model designed to match the performance of top Chinese competitors while using significantly less compute.
Efficiency over raw power
The company states Beam matches GLM 5.2 on key benchmarks while consuming three to four times less compute. It targets businesses using AI for coding and automated workflows that need to control operating costs.
Beam is a mixture-of-experts model. It activates 23 billion of its 501 billion parameters per token.
On coding and agent benchmarks, it comes close to the larger Qwen 3.8-Max. Stronger open models like Kimi K3 still beat it on raw performance. Reflection says it is already training a successor to close the gap.
Beam is text-only. However, the company says it can process content from other media formats as long as they are represented as text.
Agentic coding benchmarks
The following table compares Beam against other models on specific coding tasks:
Agentic Coding Benchmarks
- DeepSWE v1.1: Beam 44.4, GLM 5.2 44.0, GLM 5.3 61.0, Kimi K3 68.0, Qwen 3.8 Max 74.2, DeepSeek V4.1 Flash 74.2.
- SWE Bench Pro v2-Hard: Beam 77.2, Inkling 56.9, GLM 5.3 84.3, Kimi K3 88.2.
- SWE Bench Pro v1: Beam 65.5, Inkling 54.3, Nemotron 3 Ultra 46.4, GLM 5.2 62.1, DeepSeek V4.1 Flash 67.7.
- Terminal Bench v2.1: Beam 80.1, Inkling 63.8, Nemotron 3 Ultra 56.4, GLM 5.2 81.0, GLM 5.3 88.2, Kimi K3 88.3, Qwen 3.8 Max 86.6, DeepSeek V4.1 Flash 90.6.
- SWE Atlas Codebase QnA: Beam 34.6, GLM 5.3 61.0, Kimi K3 68.0.
- SWEBench Multilingual: Beam 78.0, Nemotron 3 Ultra 67.7.
- SWEBench Verified: Beam 80.9, Inkling 77.6, Nemotron 3 Ultra 70.7.
Training scale and emergent abilities
Reflection trained Beam with reinforcement learning on 10,500 Nvidia GB300 GPUs over four weeks. The company calls this one of the largest training runs any open lab has done.
Performance kept improving through the end of the run without hitting a ceiling.
Users can control how thoroughly Beam reasons through a problem. A tunable parameter lets you choose whether the model answers quickly or takes more time to think on harder tasks, trading off compute cost against output quality.
During training, the company observed what it calls “emergent capabilities”. While running an RL mix of reasoning, software engineering, and terminal tasks, Reflection noticed Beam getting better at web browsing even though no browsing tasks were part of that training mix. With web access, the model independently learned to query other language models and pull documents from external services.
Demos include a live-updating New York City subway map, a small 3D game, and a notebook for fine-tuning another AI model.
Safety and alignment
Reflection trained a second model for safety and alignment and merged it with Beam. The guidelines range from hard rules the model must never break to quality standards like factual accuracy and admitting uncertainty, along with a direct, thorough, and proactive response style.
The company plans to publish its safety test results in a technical report and open-source the evaluation methods it developed.
Release details
A technical report, developer documentation, and model weights under the open Apache 2.0 license are set to ship later this month. Beam is still going through final safety testing, and an early version is available to select users for now.
Background
Reflection was founded in 2024 by former Google Deepmind researchers Misha Laskin and Ioannis Antonoglou. Laskin led reward modeling for Gemini, and Antonoglou helped build AlphaGo.
The startup launched in March 2025 with $130 million in seed funding and the goal of building superintelligence through autonomous coding. The vision was that language models could learn to act as independently as AlphaGo plays Go, using reinforcement learning on computers.
In summer 2025, the company released Asimov, an agent for analyzing large codebases. Reflection then raised $2 billion at an $8 billion valuation in October 2025, with Nvidia among the investors. The company releases its model weights but keeps training data and pipelines proprietary.
More recently, Reflection signed billion-dollar compute deals with SpaceX and cloud provider Nebius.
