Qwen · Chat and reasoning · 180B parameters · 6B active · other
Qwen3.8 Flash Next
Needs 216 GB at FP8 / INT8. The smallest real configuration today is 8× Tesla V100 32GB at Vast.ai, $0.54 per hour, about $391 a month at 730 hours.
Parameters180B6B active per token
Context256k tokensmaximum published
Checkpoint360 GBBF16, as published
Licenseothercheck before commercial use
Released—documented date
Smallest configuration · FP8 / INT8
8× Tesla V100 32GB256 GB VRAM · 84% used
- Per hour
- $0.54
- Per month
- $391
- Cheapest at
Vast.ai
Memory · 3 precisions · 5 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 |
|---|---|---|---|---|---|---|
| BF16published checkpoint | 360 GB | 432 GB | 433 GB | 436 GB | 445 GB | 458 GB |
| FP8 / INT8estimated | 180 GB | 216 GB | 217 GB | 220 GB | 229 GB | 242 GB |
| INT4 / MXFP4estimated | 108 GB | 130 GB | 130 GB | 133 GB | 143 GB | 156 GB |
| KV cacheone sequence, FP16 cache, included above | — | — | 1 GB | 4 GB | 13 GB | 26 GB |
Cross-checked on Hugging Face Oct 28, 2026: 180.0B parameters in the safetensors headers, 360.0 GB of weight files, attention layout from config.json (48 layers, 2 KV heads × 256). 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 = 2048 bytes per token per layer × 48 full-attention layers (FP16 cache, one sequence).config.json ↗
Smallest configuration per GPU · 110 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 |
|---|---|---|---|---|---|
| Tesla V100 32GBBest fit | 8× 32 GB256 GB total | 84% | $0.54 | $391 | Vast.aiTrust 78 · read Nov 2 · receipt · also Clore.ai $0.69, Edgevana $2.21 |
| MI325X | 1× 256 GB256 GB total | 84% | $2.25 | $1,643 |  OpenRelaynot scored · read Nov 2 · receipt · also Cyfuture AI $3.14, DigitalOcean $3.80 |
| A100 80GB | 4× 80 GB320 GB total | 68% | $2.40 | $1,754 | Vast.aiTrust 78 · read Nov 2 · receipt · also Edgevana $4.60, Gcore $5.63 |
| A100 40GB | 8× 40 GB320 GB total | 68% | $3.20 | $2,338 | Vast.aiTrust 78 · read Nov 2 · receipt · also Denvr Cloud $9.20, Sesterce Cloud $12.32 |
| RTX A6000 | 8× 48 GB384 GB total | 56% | $3.36 | $2,453 |  LiumTrust 40 · read Nov 2 · receipt · also LeaderGPU $3.58, Verda $6.62 |
| L40S | 8× 48 GB384 GB total | 56% | $3.74 | $2,727 | Vast.aiTrust 78 · read Nov 2 · receipt · also Lium $6.00, GPU.ai $6.40 |
| RTX 5090 | 8× 32 GB256 GB total | 84% | $3.84 | $2,804 | Vast.aiTrust 78 · read Nov 2 · receipt · also Lium $4.00, Edgevana $5.28 |
| RTX PRO 6000 | 4× 96 GB384 GB total | 56% | $3.84 | $2,805 | Vast.aiTrust 78 · read Nov 2 · receipt · also UpCloud $3.89, Hetzner $6.44 |
| MI350X | 1× 288 GB288 GB total | 75% | $3.99 | $2,913 | OpenRelaynot scored · read Nov 2 · receipt |
| Gaudi 2 | 4× 96 GB384 GB total | 56% | $4.18 | $3,052 |  Cyfuture AInot scored · read Nov 2 · receipt |
| A40 | 8× 48 GB384 GB total | 56% | $4.61 | $3,364 |  LeaderGPUTrust 59 · read Nov 2 · receipt · also Exoscale $8.36 |
| A100X | 4× 80 GB320 GB total | 68% | $5.12 | $3,739 | Vast.aiTrust 78 · read Nov 2 · receipt |
| RTX 6000 Ada | 8× 48 GB384 GB total | 56% | $5.20 | $3,796 |  GPU.ainot scored · read Nov 2 · receipt · also io.net $5.46, QuantaCloud $6.21 |
| H200 NVL | 2× 141 GB282 GB total | 77% | $5.33 | $3,894 | Vast.aiTrust 78 · read Nov 2 · receipt · also IONOS $6.52, Geodd $6.78 |
| B300 | 1× 288 GB288 GB total | 75% | $5.57 | $4,070 |  SeewebTrust 69 · read Nov 2 · receipt · also UpCloud $5.82, Deep Infra $5.89 |
| MI300X | 2× 192 GB384 GB total | 56% | $5.71 | $4,167 | Cyfuture AInot scored · read Nov 2 · receipt · also Hot Aisle $5.98, AtmosCompute $6.69 |
| MI355X | 1× 288 GB288 GB total | 75% | $5.99 | $4,373 |  Daytonanot scored · read Nov 2 · receipt · also OpenRelay $6.30 |
| H100 | 4× 80 GB320 GB total | 68% | $6.84 | $4,991 |  GcoreTrust 80 · read Nov 2 · receipt · also GPU.ai $7.92, Edgevana $7.98 |
| H200 | 2× 141 GB282 GB total | 77% | $7.38 | $5,387 |  Theta EdgeCloudnot scored · read Nov 2 · receipt · also Vast.ai $7.53, Canopy Wave $8.00 |
| B200 | 2× 180 GB360 GB total | 60% | $8.22 | $6,002 |  UpCloudTrust 75 · read Nov 2 · receipt · also Deep Infra $9.78, Elice AI Cloud $13.64 |
| L40 | 8× 48 GB384 GB total | 56% | $8.40 | $6,132 |  OblivusTrust 82 · read Nov 2 · receipt · also CoreWeave $10.00, AtmosCompute $12.21 |
| RTX PRO 4500 | 8× 32 GB256 GB total | 84% | $9.91 | $7,235 |  AceCloudTrust 73 · read Nov 2 · receipt · also EmpirioLabs AI $11.20 |
| RTX PRO 6000D | 4× 84 GB336 GB total | 64% | $10.56 | $7,709 |  AtmosComputenot scored · read Nov 2 · receipt |
| GB300 | 1× 288 GB288 GB total | 75% | $13.05 | $9,527 |  VerdaTrust 81 · read Nov 2 · receipt · also Fireworks AI $20.00 |
| The 24 cheapest of 28 GPU types that fit. Every GPU, any precision and context, in the finder | |||||
Throughput estimates · Oct 4, 2026Tokens per second, and what a million costs
Chat and reasoning. Speed estimated from memory bandwidth and model size, never measured. The cost per 1M output tokens is an estimate at the prices read on Oct 4, 2026; it is not recomputed here.
| GPU | Cards | Precision | tok/s per user | per 1M output tokens · estimate |
|---|---|---|---|---|
| MI300X | 2 | FP8 | 71.6 | $0.611 |
| H200 | 2 | FP8 | 91.9 | $0.774 |
| B200 | 2 | FP8 | 122 | $0.825 |
| H100 SXM | 4 | FP8 | 71.3 | $1.06 |
| RTX PRO 6000 | 4 | FP8 | 47.1 | $1.36 |
| H100 PCIe | 4 | FP8 | 44.1 | $1.50 |
| L40S | 8 | FP8 | 32.4 | $1.83 |
| A100 80GB | 8 | BF16 | 46.6 | $1.90 |
cards, precision and speed estimated from memory bandwidth and model size; cost recomputed by the site from verified prices · cheapest per output at the prices of Oct 4, 2026: MI300XWhy estimates, and their limits →
Qwen3.8 Flash Next · related