Foundation Model CompaniesOpen Source
Reflection
US lab that released Beam 1.0, an Apache-2.0 text-only sparse mixture-of-experts model.
Reflection released Beam 1.0 on 2026-10-16: an Apache-2.0 text-only sparse MoE with 501B total parameters, 23B active, and a 1M-token context window, aimed at agentic coding and reasoning. Weights shipped with day-0 support in Hugging Face Transformers, vLLM, and SGLang. No first-party token price was published with the weight release.
Why Reflection matters
Beam is a new Apache-2.0 long-context MoE that can be self-hosted. It is a weight release, not a scored ranking entry, and it is unrelated to the beam.cloud GPU tool.
Last reviewed: 18 October 2026
Best fit for
When architects typically choose Reflection.
- Self-hosted coding agents
- Apache-2.0 long-context text models
Strengths
DataAIHub qualitative scores from 1 to 5, for architects. Not official benchmark results or a public ranking.
- Open weights★★★★★
- Context length★★★★☆
- Ecosystem maturity★★☆☆☆
Quick facts
- Open source
- Yes
- Flagship model
- Beam 1.0
On DataAIHub
- 1Models
- 1Products
- 2Research
- 8Guides
Company profile
- Country
- United States
- Primary focus
- Open-weight language models, Agentic coding
- Target users
- Self-hosting teams, Developers
- Deployment
- Apache-2.0 weights via Hugging Face, vLLM, and SGLang
- Licensing
- Apache-2.0 for Beam 1.0
- Open source
- Yes
- Cloud provider
- No
- Website
- https://reflection.ai
Products
Foundation models
Related research
- Introducing Beam
Original Paper
- Beam: an open large language model
Original Paper
Related guides
Explore more companies
- MetaUnited StatesMeta advances open multimodal research and releases the Llama open-weight model family used widely in self-hosted AI stacks. Muse Wonder (muse-wonder-1.0, 2026-10-12) is the newer paid Meta Model API peak for longer agentic work, with Muse Code. Muse Spark 1.3 remains the lower-priced closed API. Muse Spark Open (2026-10-15) is a separate weight release: 816B total / 42B active, 1M context, multimodal, under the Muse Model License with limited commercial terms. Combined with PyTorch, Llama still lets organizations train, fine-tune, and serve models without depending on a proprietary API.
- AlibabaChinaGlobal technology group whose Qwen team releases competitive multilingual and multimodal foundation models for cloud and open-weight use.
- DeepSeekChinaDeepSeek releases competitive open-weight models—including DeepSeek V4.1-Flash (live Flash, native multimodal), V4-Pro-0813 (still on deepseek-v4-pro after 2026-09-14), DeepSeek V3, and the reasoning-focused DeepSeek R1—plus APIs, chat, and DeepSeek Harness (MIT agent runtime, developer preview). It forced the industry to revisit cost/performance assumptions for frontier-class open models, especially for reasoning and coding workloads.
- Mistral AIFranceMistral AI builds Mistral Large 3 (with Mistral Large 4, a 1.05T-parameter multimodal MoE with 52B active parameters, in public preview since 2026-10-06), Small/Medium commercial tiers, and open-weight models, plus La Plateforme APIs and Le Chat. Mixtral remains an Apache-2.0 self-host option; it is not listed on Mistral serverless API pricing as of 2026-08-16. Regional Endpoints (EU or US) are generally available. Series D (€3B, 2026-09-08) funds sovereign open-weight infrastructure; a 2026-09-10 Cloudera partnership targets hybrid/air-gapped inference and custom models on customer data. Firefox Smart Window (beta, 2026-09-16) uses Mistral models in France and North America with zero data retention.
- Moonshot AIChinaMoonshot AI builds the Kimi product family and releases large open-weight models. Kimi K3 (2.8T MoE, 1M context) is its flagship open multimodal agentic release under the Kimi K3 License, with API access at platform.kimi.ai and hosted access on Databricks Unity AI Gateway (2026-08-06).
- Sarvam AIIndiaSarvam AI builds open-weight language models trained in India, plus speech, vision, and enterprise products aimed at Indian languages. Sarvam-M (May 2025) is a 24B Apache-2.0 model post-trained from Mistral Small. Sarvam 30B and Sarvam 105B (open-sourced 6 Mar 2026) are reasoning MoE models trained from scratch on IndiaAI compute: 30B powers the Samvaad conversational platform, and 105B powers the Indus assistant. Weights are on Hugging Face and the IndiaAI AI Kosh; the API is OpenAI-compatible.