DeepSeek · Chat and reasoning · 284B parameters · mit
DeepSeek V4 Flash
Needs 192 GB at INT4 / MXFP4. The smallest real configuration today is 8× Tesla V100 32GB at Vast.ai, $0.54 per hour, about $391 a month at 730 hours.
Parameters284Bdense
Context1M tokensmaximum published
Checkpoint160 GBINT4-packed, as published
Licensemitcheck before commercial use
ReleasedApr 24, 2026documented date
Smallest configuration · INT4 / MXFP4
8× Tesla V100 32GB256 GB VRAM · 75% used
- Per hour
- $0.54
- Per month
- $391
- Cheapest at
Vast.ai
What is documented about DeepSeek V4 Flash
news checked 2026-10-27DeepSeek V4 launched 2026-04-24: Flash at 284B parameters (13B active), Pro at 1.6T (49B active). A test run found 2× RTX PRO 6000 in tensor parallel (vLLM 0.24.0, nine workarounds needed) to be the minimum for V4 Flash in 4-bit.
Memory · 1 precision · 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 |
|---|---|---|---|---|---|---|---|
| INT4-packedpublished checkpoint | 160 GB | 192 GB | 192 GB | 195 GB | 204 GB | 215 GB | 284 GB |
| KV cacheone sequence, FP16 cache, included above | — | — | 1 GB | 3 GB | 12 GB | 24 GB | 93 GB |
Cross-checked on Hugging Face Oct 28, 2026: 290.9B parameters in the safetensors headers, 159.6 GB of weight files, attention layout from config.json (43 layers, 1 KV heads × 512). 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 × 43 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 | 75% | $0.54 | $391 | Vast.aiTrust 78 · read Nov 1 · receipt · also Clore.ai $0.69, Lium $4.40 |
| Quadro RTX 8000 | 4× 48 GB192 GB total | 100% tight | $1.01 | $741 | Vast.aiTrust 78 · read Nov 1 · receipt |
| RTX 3090 | 8× 24 GB192 GB total | 100% tight | $1.17 | $851 |  Clore.aiTrust 46 · read Nov 1 · receipt · also Vast.ai $1.29, Lium $2.32 |
| L40S | 4× 48 GB192 GB total | 100% tight | $1.52 | $1,110 |  LiumTrust 40 · read Nov 1 · receipt · also Vast.ai $1.87, VoltageGPU $3.04 |
| MI300X | 1× 192 GB192 GB total | 100% tight | $1.65 | $1,205 |  GPUBankTrust 93 · read Oct 25 · receipt · also OpenRelay $1.99, Runpod $2.39 |
| RTX PRO 6000 | 2× 96 GB192 GB total | 100% tight | $1.67 | $1,218 | Vast.aiTrust 78 · read Nov 1 · receipt · also UpCloud $1.94, Edgevana $3.07 |
| RTX A6000 | 4× 48 GB192 GB total | 100% tight | $1.68 | $1,226 | LiumTrust 40 · read Nov 1 · receipt · also Spheron $1.77, GPU.ai $1.88 |
| RTX PRO 5000 | 4× 48 GB192 GB total | 100% tight | $1.92 | $1,403 | Vast.aiTrust 78 · read Nov 1 · receipt · also Edgevana $4.60, AtmosCompute $5.28 |
| RTX PRO 4000 | 8× 24 GB192 GB total | 100% tight | $1.93 | $1,406 | Vast.aiTrust 78 · read Nov 1 · receipt · also Edgevana $2.15, AtmosCompute $3.47 |
| Gaudi 2 | 2× 96 GB192 GB total | 100% tight | $2.11 | $1,538 |  Cyfuture AInot scored · read Nov 1 · receipt |
| MI325X | 1× 256 GB256 GB total | 75% | $2.25 | $1,643 |  OpenRelaynot scored · read Nov 1 · receipt · also Cyfuture AI $3.14, DigitalOcean $3.80 |
| RTX 6000 Ada | 4× 48 GB192 GB total | 100% tight | $2.40 | $1,753 | Vast.aiTrust 78 · read Nov 1 · receipt · also GPU.ai $2.60, io.net $2.73 |
| A100 80GB | 4× 80 GB320 GB total | 60% | $2.40 | $1,754 | Vast.aiTrust 78 · read Nov 1 · receipt · also Spheron $4.76, Gcore $5.63 |
| A40 | 4× 48 GB192 GB total | 100% tight | $2.60 | $1,898 |  Denvr CloudTrust 67 · read Nov 1 · receipt · also Exoscale $4.18 |
| L40 | 4× 48 GB192 GB total | 100% tight | $2.65 | $1,932 |  io.netTrust 42 · read Nov 1 · receipt · also GPU.ai $3.12, QuantaCloud $3.75 |
| RTX 4090 | 8× 24 GB192 GB total | 100% tight | $2.99 | $2,184 | Vast.aiTrust 78 · read Nov 1 · receipt · also Sesterce Cloud $3.52, Spheron $3.84 |
| A100 40GB | 8× 40 GB320 GB total | 60% | $3.20 | $2,338 | Vast.aiTrust 78 · read Nov 1 · receipt · also Edgevana $4.67, Denvr Cloud $9.20 |
| RTX 5880 Ada | 4× 48 GB192 GB total | 100% tight | $3.20 | $2,339 | Vast.aiTrust 78 · read Nov 1 · receipt · also AtmosCompute $5.77 |
| MI350X | 1× 288 GB288 GB total | 67% | $3.99 | $2,913 | OpenRelaynot scored · read Nov 1 · receipt |
| RTX 5090 | 8× 32 GB256 GB total | 75% | $4.00 | $2,920 | LiumTrust 40 · read Nov 1 · receipt · also Vast.ai $4.16, Edgevana $4.30 |
| A100X | 4× 80 GB320 GB total | 60% | $5.12 | $3,739 | Vast.aiTrust 78 · read Nov 1 · receipt |
| H200 NVL | 2× 141 GB282 GB total | 68% | $5.33 | $3,894 | Vast.aiTrust 78 · read Nov 1 · receipt · also IONOS $6.52, Geodd $6.78 |
| A30 | 8× 24 GB192 GB total | 100% tight | $5.42 | $3,960 |  Leafcloudnot scored · read Nov 1 · receipt |
| B300 | 1× 288 GB288 GB total | 67% | $5.57 | $4,070 |  SeewebTrust 69 · read Nov 1 · receipt · also UpCloud $5.82, Deep Infra $5.89 |
| The 24 cheapest of 38 GPU types that fit. Every GPU, any precision and context, in the finder | |||||
DeepSeek V4 Flash · related