LLMs · DeepSeek
DeepSeek Coder 33B Instruct
- Parameters
- 33B
- Architecture
- Dense
- Max context
- 16K tokens
- License
- deepseek
- Released
- 2023-11
How to run it
Quickest · Ollama
One command downloads a ready-to-run version and starts a chat.
ollama run deepseek-coder:33bQuantized files · GGUF
Works in LM Studio, llama.cpp, Jan and Ollama. No sign-in needed.
huggingface.co/TheBloke/deepseek-coder-33B-instruct-GGUF (opens in a new tab)
ollama run hf.co/TheBloke/deepseek-coder-33B-instruct-GGUFOriginal weights · Hugging Face
Full-precision files from DeepSeek, for fine-tuning, vLLM, Transformers.
huggingface.co/deepseek-ai/deepseek-coder-33b-instruct (opens in a new tab)
Official repo · GitHub
Reference code, inference scripts and docs from DeepSeek. The weights themselves download from Hugging Face.
github.com/deepseek-ai/DeepSeek-Coder (opens in a new tab)
git clone https://github.com/deepseek-ai/DeepSeek-CoderMemory needed
Weights plus KV cache and runtime overhead. Leave a little extra free for the operating system and other apps.
| Quantization | Quality | Weights | Total at 8K context | Total at 16K context | Download |
|---|---|---|---|---|---|
Q2_K |
Low | 13.3 GB | 16.0 GB | 18.1 GB | GGUF (opens in a new tab) |
Q3_K_M |
Low | 16.3 GB | 19.0 GB | 21.1 GB | GGUF (opens in a new tab) |
Q4_K_M |
Good | 20.3 GB | 23.0 GB | 25.1 GB | GGUF (opens in a new tab) |
Q5_K_M |
High | 23.7 GB | 26.4 GB | 28.4 GB | GGUF (opens in a new tab) |
Q6_K |
High | 27.3 GB | 30.0 GB | 32.1 GB | GGUF (opens in a new tab) |
Q8_0 |
Very high | 35.3 GB | 38.0 GB | 40.1 GB | GGUF (opens in a new tab) |
F16 |
Full | 66.7 GB | 69.4 GB | 71.4 GB | Original weights (opens in a new tab) |
Best hardware for DeepSeek Coder 33B Instruct
The least expensive GPU in our catalog that runs it without offloading is the AMD Radeon RX 7900 XT (launch price $899) at Q2_K.
| Grade · points | Hardware | Quant | Memory | Speed | Fit |
|---|---|---|---|---|---|
| 73 ptsGreat (A) · 73 of 100 points: Model quality 18.8/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 15.0/15 | NVIDIA GeForce RTX 5090 | Q4_K_M | 24.6 GB / 32.0 GB | 66.0 tok/s | Runs well |
| 71 ptsGood (B) · 71 of 100 points: Model quality 16.6/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 15.0/15 | NVIDIA GeForce RTX 3090 | Q3_K_M | 20.2 GB / 24.0 GB | 44.6 tok/s | Runs well |
| 71 ptsGood (B) · 71 of 100 points: Model quality 16.6/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 15.0/15 | NVIDIA GeForce RTX 4090 | Q3_K_M | 20.2 GB / 24.0 GB | 48.0 tok/s | Runs well |
| 71 ptsGood (B) · 71 of 100 points: Model quality 16.6/45, Memory fit 25.0/25, Speed 14.9/15, Memory use 15.0/15 | AMD Radeon RX 7900 XTX | Q3_K_M | 20.2 GB / 24.0 GB | 39.3 tok/s | Runs well |
| 67 ptsGood (B) · 67 of 100 points: Model quality 19.6/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 8.4/15 | NVIDIA RTX PRO 6000 Blackwell | Q6_K | 34.8 GB / 96.0 GB | 45.8 tok/s | Runs well |
| 66 ptsGood (B) · 66 of 100 points: Model quality 18.8/45, Memory fit 25.0/25, Speed 9.9/15, Memory use 12.9/15 | Apple M4 Max (64 GB) | Q4_K_M | 26.9 GB / 48.0 GB | 17.3 tok/s | Runs well |
| 65 ptsGood (B) · 65 of 100 points: Model quality 18.8/45, Memory fit 25.0/25, Speed 7.0/15, Memory use 15.0/15 | Apple M4 Pro (48 GB) | Q4_K_M | 25.9 GB / 36.0 GB | 8.7 tok/s | Runs well |
| 65 ptsGood (B) · 65 of 100 points: Model quality 18.8/45, Memory fit 25.0/25, Speed 8.4/15, Memory use 12.9/15 | Apple M1 Max (64 GB) | Q4_K_M | 26.9 GB / 48.0 GB | 12.7 tok/s | Runs well |
| 61 ptsGood (B) · 61 of 100 points: Model quality 18.8/45, Memory fit 25.0/25, Speed 11.9/15, Memory use 5.5/15 | Apple M2 Ultra (192 GB) | Q4_K_M | 34.5 GB / 144 GB | 25.4 tok/s | Runs well |
| 61 ptsGood (B) · 61 of 100 points: Model quality 18.8/45, Memory fit 25.0/25, Speed 9.9/15, Memory use 7.4/15 | Apple M4 Max (128 GB) | Q4_K_M | 30.7 GB / 96.0 GB | 17.3 tok/s | Runs well |
Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.