LLMS • CURATED • UPDATED OCT 7, 2026

Reflection Beam

A 501B open-weight coding and agent model from a US lab, under Apache 2.0
Score
17.7 · rank 40
Price
Free
API
No
Weights
Yes

Beam is the first model from Reflection AI, the Nvidia-backed New York lab founded by former Google DeepMind researchers Misha Laskin and Ioannis Antonoglou. Announced October 5, 2026, it is a sparse mixture-of-experts model with 501 billion total parameters and 23 billion active, a 1-million-token context, and text-only input and output. Reflection reports 44.4% on DeepSWE v1.1, 80.1% on Terminal Bench v2.1, 80.9% on SWE-bench Verified and 90.5% on GPQA Diamond, broadly level with Z.ai's GLM-5.2 and behind Qwen3.8 Max and Kimi K3, and claims it reaches GLM-5.2's scores with 3 to 4 times less inference compute. The weights went up on Hugging Face on October 16 under Apache 2.0, in BF16, FP8 and NVFP4, with day-0 support in Transformers, vLLM and SGLang, a Rust-based agent harness, and fine-tuning and RL recipes.

The US has had few serious open-weight models since Llama stalled, and Beam is a credible one under a fully permissive licence. It does not lead its Chinese peers, and Reflection's own tables say so. It is listed for the efficiency claim and the Apache 2.0 licence, ranked below the open models it trails until it is independently tested.

Pricing Free
Platforms api, local
API No
Open Source Yes
Modalities LLMs, IDEs & Coding Tools
Best For Best US Open-Weight Contender
Date Added 2026-10-07
  • ✓501B MoE with only 23B active parameters
  • ✓Weights released under Apache 2.0 in BF16, FP8 and NVFP4
  • ✓Claims GLM-5.2-level scores at 3 to 4 times less inference compute
  • ✓1-million-token effective context
  • ✓80.9% SWE-bench Verified and 80.1% Terminal Bench v2.1, per Reflection
  • ⚠Storing the full model still needs substantial GPU memory
  • ⚠Text only
  • ⚠Trails Qwen3.8 Max and Kimi K3 on most of Reflection's own tables
  • ⚠No independent benchmark results yet

Download the weights from Hugging Face and serve them with vLLM or SGLang, using the FP8 or NVFP4 checkpoint to cut memory. Run it inside Reflection's Rust harness for agent work, and compare it on your own tasks against GLM-5.2 and Kimi K3, the open models Reflection benchmarks it against.

Artificial Analysis Intelligence Index v5.0.1 · llm
17.7
Rank #40 · 2026-11-21
DeepSWE v1.1 (Reflection-published) · ide-coding
44.4%
Free Completely free
View Reflection Beam Alternatives (2026) →

Compare Reflection Beam with 5+ similar llms AI tools.

❓
Q

Is Reflection Beam free?

A

Yes, Reflection Beam is completely free to use.

Q

Does Reflection Beam have an API?

A

No, Reflection Beam does not appear to offer a public API.

Q

What is Reflection Beam best for?

A

Reflection Beam is best for Best US Open-Weight Contender. Beam is the first model from Reflection AI, the Nvidia-backed New York lab founded by former Google DeepMind researchers Misha Laskin and Ioannis Antonoglou. The US has had few serious open-weight models since Llama stalled, and Beam is a credible one under a fully permissive licence. It does not lead its Chinese peers, and Reflection's own tables say so. It is listed for the efficiency claim and the Apache 2.0 licence, ranked below the open models it trails until it is independently tested.

Q

What platforms does Reflection Beam support?

A

Reflection Beam supports api, local.

Q

Is Reflection Beam open source?

A

Yes, Reflection Beam is open source. You can access the source code, contribute, and deploy it on your own infrastructure.

Q

How do I get started with Reflection Beam?

A

Download the weights from Hugging Face and serve them with vLLM or SGLang, using the FP8 or NVFP4 checkpoint to cut memory. Run it inside Reflection's Rust harness for agent work, and compare it on your own tasks against GLM-5.2 and Kimi K3, the open models Reflection benchmarks it against.

Q

How do I use Reflection Beam?

A

Reflection Beam is a large language model for text generation, analysis, and conversation. Use the API for programmatic access. Enter prompts or questions to get responses. It excels at 501b moe with only 23b active parameters.

🏷️

Work on Reflection Beam? You're hand-reviewed in our directory. Add this badge to your site — it links back to this profile.

Featured on CuratedAI