Reflection AI · Chat and reasoning · 501B parameters · 23B active · Apache-2.0
Beam 1.0 501B-A23B
Needs 602 GB at FP8 / INT8. The smallest real configuration today is 8× RTX PRO 6000 at UpCloud, $7.78 per hour, about $5,676 a month at 730 hours.
Parameters501B23B active per token
Context1M tokensmaximum published
Checkpoint1,003 GBBF16, as published
LicenseApache-2.0check before commercial use
ReleasedOct 16, 2026documented date
Smallest configuration · FP8 / INT8
8× RTX PRO 6000768 GB VRAM · 78% used
- Per hour
- $7.78
- Per month
- $5,676
- Cheapest at
UpCloud
What is documented about Beam 1.0 501B-A23B
news checked 2026-10-27MAJOR NEW RELEASE: Reflection AI's first open-weight model, announced 2026-10-05; weights went to HF on 2026-10-16 (Reflection blog 'Beam is now available', HF first commit). Text-only. 501B total, 23B active, 1M context. 256 experts, 8 active. 43 sliding + 9 full-attention layers. A DFlash2 drafter is also published.
- https://huggingface.co/reflection/Beam-1.0-501B-A23B
- https://huggingface.co/reflection/Beam-1.0-501B-A23B/raw/main/config.json
- https://huggingface.co/api/models/reflection/Beam-1.0-501B-A23B?blobs=true
Memory · 3 precisions · 6 context lengthsHow much VRAM it needs
Weights plus runtime overhead, then the KV cache that grows with the context you serve.
| Precision | Weights | Weights + headroom | 4k context | 32k context | 128k context | 256k context | 1M context |
|---|---|---|---|---|---|---|---|
| BF16published checkpoint | 1,003 GB | 1,204 GB | 1,204 GB | 1,205 GB | 1,209 GB | 1,213 GB | 1,242 GB |
| FP8 / INT8estimated | 501 GB | 602 GB | 602 GB | 603 GB | 607 GB | 611 GB | 640 GB |
| INT4 / MXFP4estimated | 301 GB | 361 GB | 361 GB | 363 GB | 366 GB | 371 GB | 400 GB |
| KV cacheone sequence, FP16 cache, included above | — | — | 1 GB | 2 GB | 5 GB | 10 GB | 39 GB |
Cross-checked on Hugging Face Oct 28, 2026: 501.3B parameters in the safetensors headers, 1002.7 GB of weight files, attention layout from config.json (52 layers, 8 KV heads × 128, 43 sliding-window layers). Method: weights as published at the native precision, verified on Hugging Face 2026-10-28; other precisions estimated at 1 (8-bit) and 0.6 (4-bit) bytes per parameter; × 1.2 runtime headroom; KV cache = 4096 bytes per token per layer × 9 full-attention layers + 43 sliding-window layers capped at 512 tokens (FP16 cache, one sequence).config.json ↗
Smallest configuration per GPU · 109 GPUs · cheapest eligible on-demand offer per GPU count, every provider on fileWhat it costs to run, GPU by GPU
The fewest GPUs of one type whose total memory fits the model at short context, priced from the cheapest eligible on-demand offer for exactly that count (45-day window, newest reading per provider), with the next providers beside it. Monthly = 730 hours.
| GPU | Smallest config | VRAM used | $ / hour | $ / month | Cheapest at |
|---|---|---|---|---|---|
| RTX PRO 6000Best fit | 8× 96 GB768 GB total | 78% | $7.78 | $5,676 | UpCloudTrust 75 · read Nov 2 · receipt · also Vast.ai $9.58, Lium $12.80 |
| Gaudi 2 | 8× 96 GB768 GB total | 78% | $8.27 | $6,036 |  Cyfuture AInot scored · read Nov 2 · receipt |
| A100 80GB | 8× 80 GB640 GB total | 94% tight | $8.54 | $6,231 |  Vast.aiTrust 78 · read Nov 2 · receipt · also Edgevana $9.82, GPU.ai $10.00 |
| MI300X | 4× 192 GB768 GB total | 78% | $11.31 | $8,257 | Cyfuture AInot scored · read Nov 2 · receipt · also Hot Aisle $11.96, AtmosCompute $13.38 |
| MI325X | 4× 256 GB1024 GB total | 59% | $12.29 | $8,972 | Cyfuture AInot scored · read Nov 2 · receipt |
| H100 | 8× 80 GB640 GB total | 94% tight | $13.67 | $9,982 |  GcoreTrust 80 · read Nov 2 · receipt · also Airon AI $15.04, UpCloud $16.00 |
| B200 | 4× 180 GB720 GB total | 84% | $16.46 | $12,013 | UpCloudTrust 75 · read Nov 2 · receipt · also Deep Infra $19.56, Lambda $27.16 |
| H100 NVL | 8× 94 GB752 GB total | 80% | $16.48 | $12,030 |  GPU.ainot scored · read Nov 2 · receipt · also Spheron $21.97, AtmosCompute $35.73 |
| H200 | 8× 141 GB1128 GB total | 53% | $21.04 | $15,359 |  Airon AInot scored · read Nov 2 · receipt · also Geodd $21.76, Lium $23.92 |
| B300 | 4× 270 GB1080 GB total | 56% | $23.29 | $17,004 | UpCloudTrust 75 · read Nov 2 · receipt · also FPT AI Factory $27.96, Verda $45.40 |
| MI355X | 8× 288 GB2304 GB total | 26% | $23.60 | $17,228 |  TensorWaveTrust 64 · read Nov 2 · receipt |
| H200 NVL | 8× 141 GB1128 GB total | 53% | $26.09 | $19,042 |  IONOSnot scored · read Nov 2 · receipt · also Geodd $27.12, Vast.ai $39.47 |
| B200 CC | 4× 180 GB720 GB total | 84% | $36.19 | $26,419 |  VerdaTrust 81 · read Nov 2 · receipt |
| B300 CC | 4× 270 GB1080 GB total | 56% | $46.31 | $33,806 | VerdaTrust 81 · read Nov 2 · receipt |
| A800 80GB PCIe | 8× 80 GB640 GB total | 94% tight | $50.80 | $37,088 |  Volcano Enginenot scored · read Nov 2 · receipt |
| GB300 | 4× 279 GB1116 GB total | 54% | $52.21 | $38,113 | VerdaTrust 81 · read Nov 2 · receipt |
Beam 1.0 501B-A23B · related