Reflection Beam 1.0: Pricing, Specs and Limits
Reflection Beam 1.0: open-weight sparse mixture-of-experts reasoning model, 1M context, 0.30 in / 1.00 out per 1M tokens. Specs, limits and test prompts.
By ShilpBot Team · Published · 3 min read
Beam 1.0 is Reflection's open-weight reasoning model for coding and long-horizon agentic workflows. It uses a sparse mixture-of-experts architecture with 501 billion total and 23 billion active parameters per token.
It is priced at $0.30 per million input tokens and $1.00 per million output tokens, which is low for a model of this size, and it is one of the few large open-weight options with a 1M-token window.
Prices and specifications checked against the public OpenRouter catalogue on 2026-11-16. See the live model page for current limits.
Reflection Beam 1.0 quick facts
| Detail | Value |
|---|---|
| Developer | Reflection |
| Model ID | reflection/beam-1.0-501b-a23b |
| Added to the catalogue | 2026-10-16 |
| Type | Open-weight sparse mixture-of-experts reasoning model |
| Parameters | 23B active / 501B total |
| Context window | about 1M tokens (1,048,576) |
| Maximum output | 943,718 tokens |
| Input types | text |
| Price | 0.30 in / 1.00 out per 1M tokens (cached input 0.040) |
Feature support from the catalogue: Tool calling: listed; JSON / structured output: listed; Adjustable reasoning effort: listed.
What Reflection Beam 1.0 is good for
- Coding agents and long-running workflows, which the listing targets.
- Teams that value open weights for auditing or future self-hosting.
- Long outputs, with a cap of 943,718 tokens.
Limits of Reflection Beam 1.0
- Text only. Use another model for screenshots.
- Open weights do not make the hosted endpoint private.
- A new lab, so independent track record is limited.
How Reflection Beam 1.0 compares
- MoonshotAI Kimi K3.1 Pro: Both have a 1,048,576-token context window; both cap output at 943,718 tokens; it does not accept image or video input; it costs 0.30 input and 1.00 output per million tokens against 0.24 and 8.40.
- Meta Muse Spark Open: Both have a 1,048,576-token context window; both cap output at 943,718 tokens; it does not accept image or video or file or audio input; it costs 0.30 input and 1.00 output per million tokens against 1.25 and 4.25.
- Cohere North Mini Code: It has a larger context window (1,048,576 against 256,000 tokens); its output cap is 943,718 tokens against 64,000; it activates 23B parameters per token against 3B; it is a paid model, while Cohere North Mini Code has a free listing.
Calling Reflection Beam 1.0 from code
Use the model ID reflection/beam-1.0-501b-a23b with any OpenAI-compatible client pointed at OpenRouter:
curl https://openrouter.ai/api/v1/chat/completions \
-H "Authorization: Bearer $OPENROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "reflection/beam-1.0-501b-a23b", "messages": [{"role": "user", "content": "Explain what you can do in two sentences."}]}'Prompts that show what Reflection Beam 1.0 can do
- Ask for a multi-file refactor plan and check it against your repository.
- Run a long agent task and note when it drifts from the goal.
- Compare cost and quality with Particle-1 and Kimi K3.1 Pro.
What Reflection Beam 1.0 costs
The catalogue lists 0.30 per million input tokens and 1.00 per million output tokens, with cached input at 0.040 per million. Sending 1 million input tokens and receiving 100,000 output tokens would cost about 0.40. A single coding-agent step with a 150,000-token prompt and a 10,000-token answer would cost about $0.055 before caching. If your client reuses the same prompt prefix, cached input at 0.040 per million tokens can cut the input side of that bill by about 87 percent. Prices are set by the provider, can change, and may differ between hosting providers, so check the live model page before you budget.
Keep reading
Compare every paid model by price in our AI model pricing roundup, or see the free models roundup. New to terms like mixture-of-experts? Start with how neural networks learn.
Frequently asked questions
What does 501B total, 23B active mean?
Only about 23 billion of its 501 billion parameters run for each token, which keeps speed and cost closer to a smaller model.
What are the cost and limits of Reflection Beam 1.0?
As of 2026-11-16 the catalogue lists 0.30 per million input tokens and 1.00 per million output tokens. About 1M tokens (1,048,576) context window, up to 943,718 output tokens, model ID reflection/beam-1.0-501b-a23b. Prices can change, so check the live model page.
