DeepSeek V4.1 Flash VRAM requirements
DeepSeek V4.1 Flash needs about 474 GB of VRAM at Q4_K_M with 8K context, so it needs a 640 GB setup such as 8× H100 SXM; a 24 GB card is too small. DeepSeek V4.1 Flash has 763.2B parameters, of which about 16B are used per token; all 384 experts still have to be in memory. With the same 8K context and one request it needs 783 GB at FP8 and 1,565 GB at FP16/BF16. The published weights take 475 GB (5.3 bits/weight). It does not fit on a 24 GB card at Q4_K_M; the smallest setup here that holds it with an 8K context is 8× H100 SXM (640 GB).
474 GB at Q4_K_M, 8K context, one request
- FP8
- 783 GB
- FP16 / BF16
- 1,565 GB
- Published weights
- 475 GB
- Smallest setup, Q4_K_M
- 8× H100 SXM (640 GB)
Worked example: DeepSeek V4.1 Flash with 32K tokens
Inputs: DeepSeek V4.1 Flash · Q4_K_M weights · 32,768 tokens of context · one request · FP16 KV cache
- Weights (763.2B at Q4_K_M)
- 430 GB
- KV cache (80 KB per token)
- 2.50 GB
- Buffers and runtime (0.5 GB + 10%)
- 43.8 GB
- Total
- 476 GB
The smallest setup here that holds it is 8× H100 SXM (640 GB): split across the 8 cards, with the 0.5 GB runtime base on each of them, it takes 480 GB and leaves about 160 GB to spare. Change the inputs in the calculator
What makes DeepSeek V4.1 Flash's memory use different
As a mixture-of-experts model it reads about 9.02 GB of its 430 GB Q4_K_M weights per generated token (2%), so it writes like a much smaller model while needing memory for all of them.
Its architecture and size match DeepSeek V4.1 Flash 1017, so every memory figure on this page is the same for that model.
How much VRAM does DeepSeek V4.1 Flash need?
DeepSeek V4.1 Flash needs 474 GB at Q4_K_M, 832 GB at Q8_0 and 1,565 GB at FP16 with 8,192 tokens of context and one request. Each total below is the weights plus the FP16 KV cache and the runtime overhead (0.5 GB plus 10%), and each row opens the calculator with that setting. What the GGUF names mean.
| Precision | Weights | Total | Smallest setup |
|---|---|---|---|
| As published (5.3 bits/weight) | 475 GB | 524 GB | 8× H100 SXM (640 GB) |
| FP16 / BF16 | 1,422 GB | 1,565 GB | None listed |
| FP8 / INT8 | 711 GB | 783 GB | 8× H200 (1,128 GB) |
| INT4 (AWQ / GPTQ) | 378 GB | 417 GB | 8× H100 SXM (640 GB) |
| GGUF Q8_0 | 755 GB | 832 GB | 8× H200 (1,128 GB) |
| GGUF Q6_K | 583 GB | 642 GB | 8× H200 (1,128 GB) |
| GGUF Q5_K_M | 504 GB | 555 GB | 8× H100 SXM (640 GB) |
| GGUF Q4_K_M | 430 GB | 474 GB | 8× H100 SXM (640 GB) |
| GGUF IQ4_XS | 386 GB | 426 GB | 8× H100 SXM (640 GB) |
| GGUF Q3_K_M | 347 GB | 383 GB | M3 Ultra Mac Studio (512 GB, 384 GB usable) |
| GGUF IQ3_XXS | 293 GB | 324 GB | M3 Ultra Mac Studio (512 GB, 384 GB usable) |
| GGUF Q2_K | 298 GB | 329 GB | M3 Ultra Mac Studio (512 GB, 384 GB usable) |
Fine-tuning DeepSeek V4.1 Flash? DeepSeek V4.1 Flash VRAM for LoRA, QLoRA and full training.
DeepSeek V4.1 Flash at 8K, 32K, 128K and 1M (full) tokens of context
The tables use an independent-layer upper-bound proxy for its 40-layer shared, compressed cache. Each extra token of context adds 80 KB of FP16 cache per request. The last column is the smallest setup here, one card or a group, that holds it at Q4_K_M. How the KV cache works.
| Context | KV cache, FP16 | Total at Q4_K_M | Total at Q8_0 | Smallest setup, Q4_K_M |
|---|---|---|---|---|
| 8K tokens | 640 MB | 474 GB | 832 GB | 8× H100 SXM (640 GB) |
| 32K tokens | 2.50 GB | 476 GB | 834 GB | 8× H100 SXM (640 GB) |
| 128K tokens | 10.0 GB | 485 GB | 842 GB | 8× H100 SXM (640 GB) |
| 1M tokens | 80.0 GB | 562 GB | 919 GB | 8× H100 SXM (640 GB) |
How fast DeepSeek V4.1 Flash writes on a multi-GPU group
Tokens per second for one request with 8,192 tokens of context, estimated from memory bandwidth and the parameters read per token. No single card or machine holds it at Q4_K_M or FP8 (the precisions below), so these rows are the GPU groups here that do, in tensor parallel with their combined bandwidth. A dash means it does not fit at that precision. Try other settings in the speed calculator.
| Hardware | Bandwidth | Q4_K_M | FP8 |
|---|---|---|---|
| 8× H100 SXM | 26,800 GB/s | 209–483 | — |
| 8× H200 | 38,400 GB/s | 227–544 | 202–461 |
Longest context on a multi-GPU group
No single card or machine holds it at Q4_K_M, Q8_0 or FP8 (the precisions below), so these rows use GPU groups: how many tokens of context DeepSeek V4.1 Flash fits on each, in one tensor-parallel group with one request, an FP16 KV cache and 0.5 GB left free. "Full" means the model's whole context window fits.
| Group | Memory | Q4_K_M | Q8_0 | FP8 |
|---|---|---|---|---|
| 2× RTX 3060 12GB | 24 GB | No | No | No |
| 2× RTX 3090 | 48 GB | No | No | No |
| 2× RTX 4090 | 48 GB | No | No | No |
| 2× RTX 5090 | 64 GB | No | No | No |
| 4× RTX 3090 | 96 GB | No | No | No |
| 8× H100 SXM | 640 GB | 1M (full) | No | No |
| 8× H200 | 1,128 GB | 1M (full) | 1M (full) | 1M (full) |
Model details
- Parameters
- 763.2B (763,205,315,794)
- Experts
- 384 routed experts, all loaded
- Active per token
- 16B
- Layers
- The tables use an independent-layer upper-bound proxy for its 40-layer shared, compressed cache
- Attention cache
- 1 KV heads × 512
- Context length
- 1,048,576 tokens
- Published weights
- 475 GB (5.3 bits/weight)
- On Hugging Face
- deepseek-ai/
DeepSeek-V4.1-Flash - Reported benchmarks
- GPQA Diamond 90.9. Vendor-reported: from the maker’s own model card, not rerun here; read 2026-09-30.
Why this estimate looks this way
At Q4_K_M and an 8K-token context, DeepSeek V4.1 Flash uses 430 GB for weights, 640 MB for its FP16 KV cache and 43.6 GB for estimated runtime overhead, totaling 474 GB. The overhead is a 0.5 GB base plus 10% of weights and cache.
This is a mixture-of-experts model: all 384 experts contribute to the 763.2B parameters held in memory, even though only about 16B parameters run per token. Using only active parameters would understate VRAM.
This model shares and compresses its KV cache across layers, which the calculator does not model, so the KV cache figure is an upper bound.
The architecture and published checkpoint size come from the model's config.json and weight files. These are estimates rather than measured peak memory; inference engines can reserve extra buffers or preallocate the full configured cache.
DeepSeek V4.1 Flash next to similar-sized models
The three models here closest to it in parameter count, at Q4_K_M with 8,192 tokens of context.
| Model | Parameters | Total | KV per token | Smallest setup |
|---|---|---|---|---|
| DeepSeek V4.1 Flash | 763.2B, 16B active | 474 GB | 80 KB | 8× H100 SXM (640 GB) |
| DeepSeek V4.1 Flash 1017 | 763.2B, 16B active | 474 GB | 80 KB | 8× H100 SXM (640 GB) |
| GLM-5.3 | 753.3B, 40B active | 468 GB | 90 KB | 8× H100 SXM (640 GB) |
| GLM-5.2 | 753.3B, 40B active | 468 GB | 90 KB | 8× H100 SXM (640 GB) |
DeepSeek V4.1 Flash VRAM questions
How much VRAM do I need to run DeepSeek V4.1 Flash locally?
About 474 GB at Q4_K_M with an 8K context and one request, by the bits-per-weight estimate, and 1,565 GB at BF16. The smallest setup listed here that holds Q4_K_M is the 8× H100 SXM (640 GB).
How much more VRAM does DeepSeek V4.1 Flash need for 32K or 128K tokens of context?
Its FP16 KV cache is 640 MB at 8K, 2.50 GB at 32K, 10.0 GB at 128K for one request, so 128K adds 9.38 GB over 8K, plus 10% overhead. A q8_0 cache
Does DeepSeek V4.1 Flash need less VRAM because only 16B parameters are active?
No. All 384 experts stay loaded, so the Q4_K_M weights take 430 GB. The active parameters decide speed: each generated token reads about 9.02 GB of them.
Badge for your model card
Paste it into a Hugging Face or GitHub README; it shows the Q4_K_M total above and links to this page. Q8_0 and FP16 badges.
Numbers read from the model files on Hugging Face on . Compare it with other models in the reproducible model and GPU report.