Free · no sign-up · 18 languages
A chat built for Beam, open to everyone now
Ask anything, write code, translate or think something through. Reflection AI's Beam is not publicly available yet, so today you chat with Inkling, a comparable open-weight model. When Beam opens up, this chat switches to it.
Your chats stay in this browser.
Honest status
Which model am I talking to?
Right now: Inkling, made by Thinking Machines Lab. Beam was announced by Reflection AI on October 5, 2026, but its API is waitlisted and its weights are not out yet. We will not pretend another model is Beam. The moment Beam can be served publicly, we switch and update this page.
| Answering now Inkling | Coming when available Beam | |
|---|---|---|
| Maker | Thinking Machines Lab | Reflection AI |
| Parameters (total / active) | 975B / 41B | 501B / 23B |
| Weights license | Apache 2.0 | Apache 2.0 (promised) |
| Context window | 512K | 256K (beta API) |
What you can do
Everyday help, without an account
Code
Write, explain and debug code in any popular language, with syntax-highlighted answers you can copy.
Think it through
Pick a reasoning depth: quick answers for simple questions, deeper thinking for hard ones.
Write and edit
Draft emails, posts and documents, or tighten up text you already have.
Any language
The interface comes in 18 languages and the model replies in the language you write in.
About Beam
What is Reflection Beam?
Beam is the first open-weight model from Reflection AI, a New York lab founded by former Google DeepMind researchers. It is a Mixture-of-Experts model with 501 billion parameters, of which only 23 billion work on each token, built for coding, reasoning and agent tasks.
Reflection says Beam matches GLM 5.2 on hard reasoning while using three to four times less compute, and has promised its weights under the Apache 2.0 license later in October 2026. Press have called the company the “DeepSeek of the West”.
Full Beam guide: benchmarks, specs, release tracker- total parameters
- 501B
- active per token
- 23B
- context (beta API)
- 256K
- promised license
- Apache 2.0
FAQ
Questions about BeamAI.chat
Is this the official Beam chat?+
No. BeamAI.chat is an independent site, not run by Reflection AI. Reflection offers Beam through a waitlisted beta API on its own platform.
Am I talking to Beam right now?+
No. Beam is not publicly available yet, so answers currently come from Inkling, an open-weight model by Thinking Machines Lab. The model name is shown above the chat at all times.
Why Inkling?+
It is the closest match we can serve today: like Beam it is a large Mixture-of-Experts model from a US lab, with open weights under Apache 2.0 and a strong focus on reasoning and coding.
When will the chat switch to Beam?+
As soon as Beam can be served publicly, either through open API access or once Reflection publishes the weights, which it has promised for later in October 2026.
Is it free?+
Yes. There is no sign-up and no payment. To keep it available for everyone, there is a fair-use limit on how many messages one person can send per hour.
Where are my chats stored?+
Your chat history is saved only in your browser. Messages are sent to the model provider to generate answers and are not stored on our servers. See the privacy page for details.