DeepSeek · 9B to 32B · Mixture of experts
DeepSeek Coder V2 Lite Instruct
15.71 billion parameters in total, but only 2.74 billion are read for each token. That split is the whole point of the design: it costs the memory of a large model and the speed of a small one. It compresses its cache into a latent vector, which keeps the cache small at long context.
About DeepSeek Coder V2 Lite Instruct
DeepSeek Coder V2 Lite Instruct is a mixture-of-experts model with 15.71B parameters, 2.74B of them active for each token, built from 27 layers with a hidden size of 2048. Its repository was published on 2024-06-14. The license listed on Hugging Face is other. It compresses its cache into a latent vector, which keeps the cache small at long context.
Memory needed, by quantization
At an 8k context window. Quality is our rough ranking of how much the compression costs you: anything at or above Q5 is hard to tell apart from the original in normal use.
| Quantization | File size | KV cache | Total VRAM | Quality | What it costs you |
|---|---|---|---|---|---|
| Q8_0 | 15.55 GB | 0.24 GB | 15.91 GB | 99% | Lossless in practice. Use it when the memory is there. |
| Q6_K | 12 GB | 0.24 GB | 12.41 GB | 98% | Very close to Q8 for two thirds of the size. |
| Q5_K_M | 10.41 GB | 0.24 GB | 10.84 GB | 96% | The quality-first choice when Q6 will not fit. |
| Q4_K_M | 8.83 GB | 0.24 GB | 9.29 GB | 93% | The default. Best size-to-quality ratio for local use. |
| IQ4_XS | 7.77 GB | 0.24 GB | 8.24 GB | 91% | Importance-matrix 4-bit. Q4_K_S quality, smaller file. |
| Q3_K_M | 7.15 GB | 0.24 GB | 7.63 GB | 86% | Degradation starts to show. A way to fit one size up. |
| IQ3_XXS | 5.6 GB | 0.24 GB | 6.1 GB | 79% | Aggressive. Only worth it on very large models. |
What a longer conversation costs
Same model at Q4_K_M, only the context window changes. This model compresses its cache into a latent vector, which is why the numbers stay small at long context.
| Context | KV cache | Total VRAM | Fits in 8 GB | Fits in 12 GB | Fits in 24 GB |
|---|---|---|---|---|---|
| 4k | 0.12 GB | 9.17 GB | no | yes | yes |
| 8k | 0.24 GB | 9.29 GB | no | yes | yes |
| 16k | 0.47 GB | 9.53 GB | no | yes | yes |
| 32k | 0.95 GB | 10 GB | no | yes | yes |
| 64k | 1.9 GB | 10.95 GB | no | yes | yes |
| 128k | 3.8 GB | 12.85 GB | no | no | yes |
The "fits" columns allow for the roughly 0.8 GB Windows keeps for the desktop.
Which hardware runs DeepSeek Coder V2 Lite Instruct
90 of 118 consumer devices run it at a quantization worth using, at an 8k context window.
| Device | Memory | Verdict | Quantization | Used | Speed |
|---|---|---|---|---|---|
| GeForce RTX 5090 NVIDIA | 32 GB | Runs great | Q8_0 | 15.91 GB | 262.6 t/s |
| RTX PRO 6000 Blackwell NVIDIA | 96 GB | Runs great | Q8_0 | 15.91 GB | 262.6 t/s |
| GeForce RTX 4090 NVIDIA | 24 GB | Runs great | Q8_0 | 15.91 GB | 147.7 t/s |
| GeForce RTX 3090 Ti NVIDIA | 24 GB | Runs great | Q8_0 | 15.91 GB | 147.7 t/s |
| RTX 6000 Ada Generation NVIDIA | 48 GB | Runs great | Q8_0 | 15.91 GB | 140.7 t/s |
| GeForce RTX 3090 NVIDIA | 24 GB | Runs great | Q8_0 | 15.91 GB | 137.1 t/s |
| GeForce RTX 5090 Laptop NVIDIA | 24 GB | Runs great | Q8_0 | 15.91 GB | 131.3 t/s |
| Radeon RX 7900 XTX AMD | 24 GB | Runs great | Q8_0 | 15.91 GB | 119.3 t/s |
| RTX A6000 NVIDIA | 48 GB | Runs great | Q8_0 | 15.91 GB | 112.5 t/s |
| RTX A5000 NVIDIA | 24 GB | Runs great | Q8_0 | 15.91 GB | 112.5 t/s |
| Apple M3 Ultra 96GB Apple | 96 GB | Runs great | Q8_0 | 16.09 GB | 110.9 t/s |
| Apple M3 Ultra 256GB Apple | 256 GB | Runs great | Q8_0 | 16.09 GB | 110.9 t/s |
| Apple M3 Ultra 512GB Apple | 512 GB | Runs great | Q8_0 | 16.09 GB | 110.9 t/s |
| Apple M1 Ultra 64GB Apple | 64 GB | Runs great | Q8_0 | 16.09 GB | 108.3 t/s |
| Apple M1 Ultra 128GB Apple | 128 GB | Runs great | Q8_0 | 16.09 GB | 108.3 t/s |
| Apple M2 Ultra 64GB Apple | 64 GB | Runs great | Q8_0 | 16.09 GB | 108.3 t/s |
| Apple M2 Ultra 128GB Apple | 128 GB | Runs great | Q8_0 | 16.09 GB | 108.3 t/s |
| Apple M2 Ultra 192GB Apple | 192 GB | Runs great | Q8_0 | 16.09 GB | 108.3 t/s |
| Radeon PRO W7900 AMD | 48 GB | Runs great | Q8_0 | 15.91 GB | 107.4 t/s |
| Radeon RX 7900 XT AMD | 20 GB | Runs great | Q8_0 | 15.91 GB | 99.4 t/s |
| Apple M4 Max 36GB Apple | 36 GB | Runs great | Q8_0 | 16.09 GB | 73.9 t/s |
| Apple M4 Max 48GB Apple | 48 GB | Runs great | Q8_0 | 16.09 GB | 73.9 t/s |
| Apple M4 Max 64GB Apple | 64 GB | Runs great | Q8_0 | 16.09 GB | 73.9 t/s |
| Apple M4 Max 128GB Apple | 128 GB | Runs great | Q8_0 | 16.09 GB | 73.9 t/s |
| Apple M1 Max 32GB Apple | 32 GB | Runs great | Q8_0 | 16.09 GB | 54.2 t/s |
Show 35 more devices
| Apple M1 Max 64GB Apple | 64 GB | Runs great | Q8_0 | 16.09 GB | 54.2 t/s |
| Apple M2 Max 32GB Apple | 32 GB | Runs great | Q8_0 | 16.09 GB | 54.2 t/s |
| Apple M2 Max 64GB Apple | 64 GB | Runs great | Q8_0 | 16.09 GB | 54.2 t/s |
| Apple M2 Max 96GB Apple | 96 GB | Runs great | Q8_0 | 16.09 GB | 54.2 t/s |
| Apple M3 Max 36GB Apple | 36 GB | Runs great | Q8_0 | 16.09 GB | 54.2 t/s |
| Apple M3 Max 48GB Apple | 48 GB | Runs great | Q8_0 | 16.09 GB | 54.2 t/s |
| Apple M3 Max 64GB Apple | 64 GB | Runs great | Q8_0 | 16.09 GB | 54.2 t/s |
| Apple M3 Max 96GB Apple | 96 GB | Runs great | Q8_0 | 16.09 GB | 54.2 t/s |
| Apple M3 Max 128GB Apple | 128 GB | Runs great | Q8_0 | 16.09 GB | 54.2 t/s |
| NVIDIA DGX Spark 128GB NVIDIA | 128 GB | Runs great | Q8_0 | 16.12 GB | 40 t/s |
| Apple M4 Pro 24GB Apple | 24 GB | Just fits | Q8_0 | 16.09 GB | 37 t/s |
| Apple M4 Pro 48GB Apple | 48 GB | Runs great | Q8_0 | 16.09 GB | 37 t/s |
| Apple M4 Pro 64GB Apple | 64 GB | Runs great | Q8_0 | 16.09 GB | 37 t/s |
| Jetson AGX Orin 32GB NVIDIA | 32 GB | Runs great | Q8_0 | 16.12 GB | 30 t/s |
| Jetson AGX Orin 64GB NVIDIA | 64 GB | Runs great | Q8_0 | 16.12 GB | 30 t/s |
| Ryzen AI Max+ 395 32GB AMD | 32 GB | Runs great | Q8_0 | 16.12 GB | 29 t/s |
| Ryzen AI Max+ 395 64GB AMD | 64 GB | Runs great | Q8_0 | 16.12 GB | 29 t/s |
| Ryzen AI Max+ 395 96GB AMD | 96 GB | Runs great | Q8_0 | 16.12 GB | 29 t/s |
| Ryzen AI Max+ 395 128GB AMD | 128 GB | Runs great | Q8_0 | 16.12 GB | 29 t/s |
| Apple M1 Pro 32GB Apple | 32 GB | Runs great | Q8_0 | 16.09 GB | 27.1 t/s |
| Apple M2 Pro 32GB Apple | 32 GB | Runs great | Q8_0 | 16.09 GB | 27.1 t/s |
| Apple M5 24GB Apple | 24 GB | Just fits | Q8_0 | 16.09 GB | 20.7 t/s |
| Apple M5 32GB Apple | 32 GB | Runs well | Q8_0 | 16.09 GB | 20.7 t/s |
| Apple M3 Pro 36GB Apple | 36 GB | Runs well | Q8_0 | 16.09 GB | 20.3 t/s |
| GeForce RTX 5080 NVIDIA | 16 GB | Runs great | Q6_K | 12.41 GB | 178 t/s |
| GeForce RTX 5070 Ti NVIDIA | 16 GB | Runs great | Q6_K | 12.41 GB | 166.1 t/s |
| GeForce RTX 5080 Laptop NVIDIA | 16 GB | Runs great | Q6_K | 12.41 GB | 142.4 t/s |
| GeForce RTX 4080 SUPER NVIDIA | 16 GB | Runs great | Q6_K | 12.41 GB | 136.4 t/s |
| GeForce RTX 4080 NVIDIA | 16 GB | Runs great | Q6_K | 12.41 GB | 132.9 t/s |
| GeForce RTX 4070 Ti SUPER NVIDIA | 16 GB | Runs great | Q6_K | 12.41 GB | 124.6 t/s |
| GeForce RTX 4090 Laptop NVIDIA | 16 GB | Runs great | Q6_K | 12.41 GB | 106.8 t/s |
| Radeon RX 9070 XT AMD | 16 GB | Runs great | Q6_K | 12.41 GB | 101.4 t/s |
| Radeon RX 9070 AMD | 16 GB | Runs great | Q6_K | 12.41 GB | 101.4 t/s |
| Radeon RX 7800 XT AMD | 16 GB | Runs great | Q6_K | 12.41 GB | 98.1 t/s |
| Radeon RX 7900 GRE AMD | 16 GB | Runs great | Q6_K | 12.41 GB | 90.6 t/s |
How it compares
Other models in the same family or of a similar size, on a 12 GB card and a 24 GB card, each at the best quantization that fits.
| Model | Weights at 4 bits | GeForce RTX 3060 12GB | GeForce RTX 4090 |
|---|---|---|---|
| DeepSeek Coder V2 Lite Instruct | 8.8 GB | 75.7 t/s at Q5_K_M | 147.7 t/s at Q8_0 |
| DeepSeek Coder 7B Instruct V1.5 | 3.9 GB | 26 t/s at Q8_0 | 72.9 t/s at Q8_0 |
| DeepSeek Coder 6.7B Instruct | 3.8 GB | 25.1 t/s at Q8_0 | 70.4 t/s at Q8_0 |
| DeepSeek R1 | 384.9 GB | does not fit well | does not fit well |
| GPT OSS 20B | 11.3 GB | 65.7 t/s at MXFP4 | 180.1 t/s at MXFP4 |
| Gemma 4 12B Instruct | 6.7 GB | 25.4 t/s at Q6_K | 56.4 t/s at Q8_0 |
What to buy to run DeepSeek Coder V2 Lite Instruct
The cheapest hardware that runs it at a quantization worth using and a speed you would not resent, at an 8k context window.
Cheapest that works
GeForce GTX 1650
$100 used
Q4_K_M · uses 9.29 GB of its 4 GB · about 25.6 tokens/s
Best value
GeForce RTX 2060 12GB
$160 used
Q5_K_M · about 70.7 tokens/s · 336 GB/s
| Device | Price | Memory | Runs it at | Speed | |
|---|---|---|---|---|---|
| GeForce GTX 1650 NVIDIA | $100 used | 4 GB | Q4_K_M | 25.6 t/s | Check price |
| GeForce GTX 1660 SUPER NVIDIA | $110 used | 6 GB | Q4_K_M | 46 t/s | Check price |
| GeForce GTX 1080 Ti NVIDIA | $150 used | 11 GB | Q4_K_M | 117.5 t/s | Check price |
| GeForce RTX 2060 12GB NVIDIA | $160 used | 12 GB | Q5_K_M | 70.7 t/s | Check price |
| GeForce RTX 2080 Ti NVIDIA | $250 used | 11 GB | Q4_K_M | 149.6 t/s | Check price |
| Arc A770 16GB Intel | $280 used | 16 GB | Q6_K | 80.2 t/s | Check price |
| GeForce RTX 3080 10GB NVIDIA | $350 used | 10 GB | IQ4_XS | 205.9 t/s | Check price |
| Radeon RX 7900 XT AMD | $650 | 20 GB | Q8_0 | 99.4 t/s | Check price |
Indicative prices reviewed 2026-09-01; used prices are marketplace typical. Options that cost more than a cheaper one with no more memory and no more speed are hidden. Price links are Amazon affiliate links. Change the standard, the context window or the budget in the buying tool.
These numbers also cover
Fine-tunes share their base model's architecture, so they need exactly the same memory. If you are looking for one of these, the figures above apply unchanged.
deepseek-ai/DeepSeek-V2-Lite-Chatdeepseek-ai/DeepSeek-V2-Lite
Where it sits in the catalog
Source: deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct on Hugging Face. Downloaded 718k times in the last month. Published 2024-06-14. Architecture figures are read from the repository's own configuration file, so they move when the model does.
Questions people ask about DeepSeek Coder V2 Lite Instruct
How much VRAM does DeepSeek Coder V2 Lite Instruct need?
At Q4_K_M with an 8k context window, DeepSeek Coder V2 Lite Instruct needs about 9.29 GB: 8.72 GB of weights, 0.24 GB of KV cache and 0.34 GB of overhead. At the highest-quality file it needs about 15.91 GB.
What is the cheapest GPU that runs DeepSeek Coder V2 Lite Instruct?
The smallest consumer device in our catalog that fits it is the GeForce RTX 3080 10GB with 10 GB, running it at IQ4_XS at roughly 205.9 tokens per second.
How much does a longer context window add for DeepSeek Coder V2 Lite Instruct?
On the memory side, a 32k window adds 0.9 GB of KV cache on top of the weights, and the cache grows more slowly than a conventional model because of its attention design. The longest window the model supports natively is 160k tokens.