DeepSeek v2 — Hardware Requirements & GPU Compatibility
Chat
DeepSeek-V2 is DeepSeek's base pretrained Mixture-of-Experts language model, not instruction-tuned, comprising about 235.7 billion total parameters with roughly 21.4 billion activated per token. It was pretrained on 8.1 trillion tokens and built for economical training and efficient inference, cutting training costs and KV-cache size sharply compared with DeepSeek's earlier dense 67B model while boosting generation throughput. Instruction-tuned SFT and RL chat variants were released separately as DeepSeek-V2-Chat. Despite the large total parameter count, its MoE design activates only a fraction per token, but it still needs a multi-GPU server to run at full precision. Context length is 163,840 tokens. It is released under a custom DeepSeek Model License that permits free use, including commercial use, subject to acceptable-use restrictions in the license agreement. It was published in April 2024, ahead of the DeepSeek-V2-Chat and later DeepSeek-V3 models.
DeepSeek v2 runs on a 192 GB+ Mac or a datacenter GPU
Best local pick → Mac Pro M2 Ultra (192 GB) · Q4_K_M · 142.9 GB · ~28 tok/s
Macs from 128 GB (Q3_K_M · ~32 tok/s)
Other
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run DeepSeek v2
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
ollama run hf.co/mradermacher/DeepSeek-V2-i1-GGUFQ4_K_M runs on one GH200 Grace Hopper Superchip 144 GB · 16K context
Download: DeepSeek-V2-i1-Q4_K_M.gguf141.44 GBmradermacher
Ways to runOther ways
each opens step-by-stepRun in cloud
Runs on B200 (192 GB) (49 GB headroom) · Q4_K_M
- Generation speed
- ~172 tok/s
- generation speed
- Cost per 1M output tokens
- $9.66
- per 1M output tokens
How Much VRAM Does DeepSeek v2 Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | Runs on | |
|---|---|---|---|---|
| 103.0 GB | 102.8 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Runs on Instinct MI250X 128 GB | |
| 117.7 GB | 117.5 GB | 3-bit medium — 3-bit medium quantization | Runs on Instinct MI250X 128 GB | |
| 144.3 GB | 144 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Runs on GH200 Grace Hopper Superchip 144 GB | |
| 170.8 GB | 170.5 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Runs on B200 192 GB | |
| 197.3 GB | 197.1 GB | 6-bit, very good — 6-bit quantization, very good quality | Runs on Instinct MI325X 256 GB | |
| 238.6 GB | 238.3 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Runs on Instinct MI325X 256 GB | |
| 474.3 GB | 474.1 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Runs on 7× H100 SXM 80 GB |
Selected Q4_K_M
4-bit medium quantization142.9GBat 16K context
- Weights
- 141.4 GB
- KV cache
- 1.1 GB
- Overhead
- 0.3 GB
Server-class: ~142.9 GB at 16K context · one GH200 Grace Hopper Superchip 144 GB
Or .
Which Hardware Can Run DeepSeek v2?
Q4_K_M · 144.3 GBDeepSeek v2 (Q4_K_M) requires 144.3 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 188+ GB is recommended. Using the full 164K context window can add up to 8.8 GB, bringing total usage to 153.1 GB. No consumer GPU has enough memory. 4 devices with unified memory can run DeepSeek v2, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Pro M2 Ultra (192 GB).
128–144 GB
Q4_K_M · best fit Q3_K_M (116.4 GB)
4 GPUs · 2 Mac classes · 1 device family
GH200 Grace Hopper Superchip~150Simulate output on GH200 Grace Hopper SuperchipH200 NVL~149Simulate output on H200 NVLH200 SXM~149Simulate output on H200 SXMM2 Ultra*128 GB~32Simulate output on Mac Pro M2 Ultra (128 GB)M5 Max128 GB~27Simulate output on Mac Studio M5 Max (128 GB)GB10~16Simulate output on NVIDIA DGX Spark192+ GB
fits · 49.1 GB free · up to 160K ctx
6 GPUs · 1 Mac class · 2 device families
B200~172Simulate output on B200B300~172Simulate output on B300Instinct MI350X~68Simulate output on Instinct MI350XM3 Ultra*256–512 GB~29Simulate output on Mac Studio (M3 Ultra, 256GB)8x H100 SXM5~247Simulate output on NVIDIA DGX H1008x A100 SXM4~221Simulate output on NVIDIA DGX A100 640GB
By VRAM tier · typical GPUs and Macs at Q4_K_M
Consumer GPUs (up to 32 GB) and Macs up to 96 GB: too small for this model (needs ≥ 101.6 GB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →
How good is DeepSeek v2?
llmrun Score blends public benchmark boards into one number; 100 = the average open model.
Not enough data yet for an llmrun Score.
A Score needs results on enough independent boards. How the Composite Score is built
Benchmarks
The 6 boards that weigh most in the llmrun Score, of 7 with a result for this model; badges show who ran the test.
- MMLU
Epoch AI · May 2024Lab-reported
Lab-reported78.4%
#15/85 open
- BIG-Bench Hard
Epoch AI · May 2024Lab-reported
Lab-reported78.8%
#6/38 open
- HellaSwag
Epoch AI · May 2024Lab-reported
Lab-reported87.1%
#4/47 open
- ARC Challenge
Epoch AI · May 2024Lab-reported
Lab-reported92.2%
#4/56 open
- WinoGrande
Epoch AI · May 2024Lab-reported
Lab-reported86.3%
#3/51 open
- TriviaQA
Epoch AI · May 2024Lab-reported
Lab-reported80.0%
#5/19 open
Alternatives to DeepSeek v2
DeepSeek V2 family · 5 models →Similar models
5| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| MiniMax M2.7MiniMax | 107.4 | 139.6 GB | —needs > 24 GB | same size |
| Command A Plus 05 2026 BF16Cohere | — | 132.1 GB | —needs > 24 GB | same size |
| Qwen3 VL 235B A22B InstructAlibaba | — | 143.3 GB | —needs > 24 GB | same size |
| Laguna M.1poolside | — | 138.1 GB | —needs > 24 GB | same size |
| Llama 3 1 Nemotron Ultra 253B V1NVIDIA | — | 564.2 GB | —needs > 24 GB | same size |
Specifications
- Publisher
- DeepSeek
- Family
- DeepSeek V2
- Parameters
- 235.7B
- Active parameters
- 21.4B per token
- Architecture
- DeepseekV2ForCausalLM
- Context Length
- 163,840 tokens
- Vocabulary Size
- 102,400
- Release Date
- 2024-04-22
- License
- Other
Frequently Asked Questions
How much VRAM does DeepSeek v2 need?
DeepSeek v2 requires 144.3 GB of VRAM at Q4_K_M, or 474.3 GB at BF16. Full 164K context adds up to 8.8 GB (153.1 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 235.7B × 4.8 bits ÷ 8 = 141.4 GB
KV Cache + Overhead ≈ 2.9 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 11.7 GB (at full 164K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run DeepSeek v2?
No — DeepSeek v2 requires at least 103.0 GB at Q2_K, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
What's the best quantization for DeepSeek v2?
For DeepSeek v2, Q4_K_M (144.3 GB) offers the best balance of quality and VRAM usage. Q5_K_M (170.8 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 103.0 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run DeepSeek v2 on a Mac?
Yes — Mac Pro M2 Ultra (128 GB) and 14 other Macs can run DeepSeek v2, the smallest at Q3_K_M. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 144.3 GB of usable unified memory (RAM minus macOS overhead), from Mac Pro M2 Ultra (192 GB).
Can I run DeepSeek v2 locally?
Yes — DeepSeek v2 can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is DeepSeek v2?
At Q4_K_M, DeepSeek v2 can reach ~68 tok/s on AMD Instinct MI350X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)
Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.
Example: NVIDIA B200 → 13.1 GB active ÷ (8000 × 0.65) = 2.52 ms, plus 60 layers × 0.055 ms = 3.30 ms, so 1000 ÷ 5.82 ms = ~172 tok/s
Estimated speed at Q4_K_M (144.3 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of DeepSeek v2?
At Q4_K_M, the download is about 141.44 GB. The full-precision BF16 version is 471.48 GB. The smallest option (Q2_K) is 100.19 GB.
Which GPUs can run DeepSeek v2?
No single consumer GPU has enough VRAM to run DeepSeek v2 at Q4_K_M (144.3 GB). Multi-GPU or professional hardware is required.
Which devices can run DeepSeek v2?
4 devices with unified memory can run DeepSeek v2 at Q4_K_M (144.3 GB), including Mac Pro M2 Ultra (192 GB), Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio M2 Ultra (192 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.