DeepSeek Coder 33B Instruct on Orange Pi AIpro (20T) (24 GB)

Can the Orange Pi AIpro (20T) (24 GB) run DeepSeek Coder 33B Instruct?

Barely: it runs very slowly (tight fit)

Very slow (tight fit). At Q3_K_M it needs about 20.7 GB of 21.1 GB usable, with an estimated speed of 0.7 tok/s.

Best quant: Q3_K_M
Generation speed: 0.7 tok/s
Prompt processing: 0.9 tok/s
First token (1K prompt): 1,138.0 s
Largest safe context: 9K

Best app to run it on the Orange Pi AIpro (20T) (24 GB)

On Linux on ARM boards

Ollama

Command line + local API · runs on llama.cpp (ggml), MLX (Apple silicon, preview)

One command with Ollama’s own build of this model.
Serves an OpenAI-compatible API that other apps and coding tools can use.

ollama run deepseek-coder:33b-instruct-q3_K_M

LM Studio Point-and-click: search, download and chat in one window.
llama.cpp Usually the first to support new model architectures.

Want an agent? DeepSeek Coder 33B Instruct supports tool use, so you can pair a llama.cpp or MLX LM server with one of these. Apps that run on this computer come first.

Coding agents

Personal assistants

How we choose: the files published for this model, your operating system and GPU, whether it fits fully in memory, and published speed tests (MLX vs llama.cpp on Apple silicon, vLLM vs llama.cpp vs Ollama on NVIDIA). Details.

Download DeepSeek Coder 33B Instruct

Recommended for the Orange Pi AIpro (20T) (24 GB): Q3_K_M. Every other quantization has its own download link in the table below.

Quickest · Ollama

One command downloads a ready-to-run version and starts a chat.

Get Ollama at ollama.com/download (opens in a new tab) · model page (opens in a new tab)

ollama run deepseek-coder:33b-instruct-q3_K_M

Quantized files · GGUF

Works in LM Studio, llama.cpp, Jan and Ollama. No sign-in needed.

Download the Q3_K_M file (opens in a new tab)

huggingface.co/TheBloke/deepseek-coder-33B-instruct-GGUF (opens in a new tab)

ollama run hf.co/TheBloke/deepseek-coder-33B-instruct-GGUF:Q3_K_M

Original 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)

No sign-in needed.

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-Coder

Links checked 3 days ago against Hugging Face, Ollama and GitHub.

How it’s rated

Quality 36Basic quality (D) · 36 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 57/100 (quality, fit, speed and memory use) sets the ranking.: How capable the model itself is at Q3_K_M, from its editorial quality tier scaled by the quantization or precision. It’s the same on any hardware that runs this variant. Scale: A+ 85+ · A 70+ · B 55+ · C 40+ · D under 40.
Runs Very slowtight fit: How well it runs on the Orange Pi AIpro (20T) (24 GB): speed, capped by how well it fits in memory. Runs great: 30+ tokens/s (or near the image, video or decision speed target) · Runs well: 10+ tokens/s · Slow: 3+ tokens/s · Very slow: under 3 · Won’t run: doesn’t fit even with offload. A tight fit or RAM offload caps it at Runs well; heavy offload caps it at Slow.

Match score

To order lists, each result also gets a match score out of 100 from four parts. It favors the most capable model that still runs well.

Model qualityEditorial quality tier of the model, scaled by the quantization or precision used.	16.6 / 45
Memory fitFull marks when the model fits in the share of unified memory the GPU can use; fewer when it goes past that limit.	24.0 / 25
SpeedGeneration speed against a 40 tokens/s target for LLMs, or render time against a target for image and video.	2.0 / 15
Memory useRewards using the memory you have, so a bigger, better variant beats a tiny one that leaves most memory idle.	15.0 / 15
Match score	57.6 / 100

The match score orders results; it is not shown as a grade. Quality and Runs are the two ratings in the tables.

Every quantization on the Orange Pi AIpro (20T) (24 GB)

Quant	Memory needed	Available	Speed	Runs	Quality	Download
`Q2_K`	17.7 GB	21.1 GB	0.9 tok/s	Very slow	30Basic quality (D) · 30 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 56/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q3_K_M`	20.7 GB	21.1 GB	0.7 tok/s	Very slowtight fit	36Basic quality (D) · 36 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 57/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q4_K_M`	24.7 GB	21.1 GB	—	Won't run	41Fair quality (C) · 41 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 5/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q5_K_M`	28.0 GB	21.1 GB	—	Won't run	42Fair quality (C) · 42 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 5/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q6_K`	31.7 GB	21.1 GB	—	Won't run	43Fair quality (C) · 43 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 5/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q8_0`	39.7 GB	21.1 GB	—	Won't run	44Fair quality (C) · 44 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 5/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`F16`	71.0 GB	21.1 GB	—	Won't run	44Fair quality (C) · 44 of 100: how capable the model is at F16, whatever it runs on. Match score 5/100 (quality, fit, speed and memory use) sets the ranking.	Original weights (opens in a new tab)

Memory figures assume 8K of context. Longer conversations or documents grow the KV cache.

Other hardware that runs it

41Fair quality (C) · 41 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 73/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 5090 Q4_K_M · 66.0 tok/s · Runs great
36Basic quality (D) · 36 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 71/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 3090 Q3_K_M · 44.6 tok/s · Runs great
36Basic quality (D) · 36 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 71/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 4090 Q3_K_M · 48.0 tok/s · Runs great
36Basic quality (D) · 36 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 71/100 (quality, fit, speed and memory use) sets the ranking. AMD Radeon RX 7900 XTX Q3_K_M · 39.3 tok/s · Runs great
43Fair quality (C) · 43 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 67/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA RTX PRO 6000 Blackwell Q6_K · 45.8 tok/s · Runs great
41Fair quality (C) · 41 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 66/100 (quality, fit, speed and memory use) sets the ranking. Apple M4 Max (64 GB) Q4_K_M · 17.3 tok/s · Runs well

Quick answers

How much memory does DeepSeek Coder 33B Instruct need on the Orange Pi AIpro (20T) 24GB?

At Q3_K_M it needs about 20.7 GB. The Orange Pi AIpro (20T) 24GB has 21.1 GB usable.

How fast is DeepSeek Coder 33B Instruct on the Orange Pi AIpro (20T) 24GB?

About 0.7 tok/s at Q3_K_M, an estimate from the hardware’s memory bandwidth and the size of the model.

Which quantization of DeepSeek Coder 33B Instruct should I use on the Orange Pi AIpro (20T) 24GB?

Q3_K_M. Of the versions that fit, it is the one that runs best; when two run equally well, the check prefers the higher-rated one.

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.