Current model · Reflection AI
Beam 1.0 benchmarks, pricing and specs
Beam 1.0 is a Reflection AI model released on October 16, 2026 that takes text input and produces text. Its price is not published and accepts up to 1,000,000 tokens of context.
- Context window
- 1M#26 of 64 current models
Data verified October 23, 2026 · Dataset 2026-10-23
Specs
Key facts
- Provider
- Reflection AI
- Model family
- Beam
- Release date
- October 16, 2026
- Status
- Current
- Access
- API and open weights
- API model id
- Beam-1.0-501B-A23B
- Context window
- 1,000,000 tokens
- Max output
- Not reported
- Input
- text
- Output
- text
- Task families
- text
- Declared capabilities
- reasoning, coding, agentic coding
Charts
Against every current text model
Where Beam 1.0 lands among the current models of its task, on every independent benchmark, on price and on context.
Where it stands
Beam 1.0 is no longer a current model, so it is not ranked; each bar is the share of today's text models its value beats. Each bar is the share of current models it beats.
Price and context
Context window
#26of 64
1M tokens · longer than 22%
Benchmarks
Independent benchmark results
Scores come from independent evaluators that ran the model themselves, pinned to a dated snapshot. Rank is among current models in the same snapshot. Reflection AI's own launch figures are listed under strengths with the publisher named, because they are not measured the same way.
No independent benchmark has published a result for Beam 1.0 yet. It is still compared below on price, context and capabilities.
Fit
Strengths, cautions and best uses
- Strengths
- A 501B sparse mixture-of-experts model with 23B active parameters and a 1M-token context, built for coding, reasoning and agents
- Apache 2.0 weights in BF16, FP8 and NVFP4, with no ban on distillation, fine-tuning or training on its outputs
- Reflection reports 81.6% on SWE-bench Verified and 90.6% on GPQA Diamond for the BF16 weights
- Cautions
- Reflection API is a public beta that is free to start, with no published per-token price
- Text only, with no image, audio or video input, and no independent benchmark score is recorded here yet
- Best for
- Self-hosted coding agents under a permissive licence
- Fine-tuning or distilling a large open model
From benchmark to production
Building on Beam 1.0?
Public scores rarely hold on your own data. We test it against the alternatives on your real tasks, then build the agent or pipeline around the model that wins.
Alternatives
Alternatives to Beam 1.0
The closest current text models by price. The cheapest model that scores as well or better is listed first; failing that, the cheapest one within 3% of the benchmark's scale.
Muse Spark Open
- From Meta
Nemotron 3.5 Bolt 74B-A9B
- 262K context window vs 1M
- From NVIDIA
Nemotron 3.5 Super VL
- From NVIDIA
Mercury 3
- From Inception
FAQ
Beam 1.0 FAQ
What is the context window of Beam 1.0?
Beam 1.0 accepts up to 1,000,000 tokens of context. That ranks #26 of 64 current text models by context length.
Is Beam 1.0 open source or API only?
Beam 1.0 is available both through Reflection AI's API and as open weights you can run yourself.
What are the best alternatives to Beam 1.0?
The closest current alternatives by price and measured performance are Muse Spark Open (from Meta), Nemotron 3.5 Bolt 74B-A9B (262K context window vs 1M), Nemotron 3.5 Super VL (from NVIDIA) and Mercury 3 (from Inception).
Sources
Sources
- Beam is now available · Reflection AI · official · verified October 23, 2026
- Beam 1.0 501B-A23B model card · Reflection AI · official · verified October 23, 2026
- Reflection developer documentation · Reflection AI · official · verified October 23, 2026
Cite this page
Van Data Team. "Beam 1.0 Benchmarks, Pricing and Specs." AI Model Benchmark Dataset 2026-10-23. https://vandatateam.com/ai-model-benchmarks/beam-1-0
Reuse the data: JSON · CSV · or explore every model in the AI model benchmark comparison.
Production evaluation
Deciding whether Beam 1.0 fits your workload?
Public benchmarks narrow the shortlist. We test the finalists on your own tasks, data and cost limits, then build the routing and review controls to run the winner in production.
- Evaluation cases and acceptance thresholds tied to real work
- Model routing, fallbacks and cost controls