Arcee AI, released 2026-01-27. Sparse MoE decoder, 400B total, 13B active, 512K context. Attention: GQA with gated attention and 3:1 sliding-window/global attention.
Sparse MoEReflection 🇺🇸 · October 16, 2026
Reflection Beam (501B)
Text-only MoE decoder with three dense FFN layers followed by 49 expert layers. Each MoE layer activates 8 of 256 experts without a shared expert.
Key detail
Combines 512-token local RoPE attention with global NoPE and query scaling. Uses sandwich RMSNorm, depth-scaled residuals, and a final logit softcap of 30.Total params
501B
Active / token
23B
Context
1M
Fact sheet
- Scale
- 501.3B total, 23.4B active (4.7% active)
- Context window
- 1,048,576 tokens
- Decoder type
- Sparse MoE
- Attention
- Gated GQA with QK-Norm and sliding-window/global attention
- Layer mix
- 43 sliding-window + 9 global GQA
- KV cache / token (bf16)
- 208 KiB · High
- License
- Apache License 2.0
- Released
Efficiency
Parameters active per token
4.7%
23B of 501B weights run for each token.
AA Intelligence Index v5.0
Source17.7
- Knowledge Work
- 26.3
- Coding
- 4.5
- Science
- 10.9
- General
- 29.2
Architecture concepts
Sources & resources
Use the ← and → keys to step through releases.
Similar architectures
Meta, released 2026-10-15. Sparse MoE decoder, 816B total, 42B active, 1M context. Attention: Gated GQA with scaleless QK-Norm, local-only RoPE, global NoPE, and 3:1 sliding-window/global attention.
StepFun, released 2026-09-20. Sparse MoE decoder, 600B total, 27B active, 1M context. Attention: Gated GQA with QK-Norm and 3:1 sliding-window/sparse attention.
Aleph Alpha, released 2026-10-03. Sparse MoE decoder, 78B total, 3.5B active, 256K context. Attention: GQA with QK-Norm and 4:1 sliding-window/global attention.