Qwen 3 235B-A22BvsDeepSeek R1 671B
Complete comparison between Qwen 3 235B-A22B (235B parameters, Alibaba) and DeepSeek R1 671B (671B, DeepSeek). VRAM required by quantization, tokens per second measured on 4 reference GPUs, verdict by use case, license, installation commands. All figures are calculated from catalog data — no copy-pasting between pages.
In brief
| Feature | Qwen 3 235B-A22B | DeepSeek R1 671B |
|---|---|---|
| Settings | 235B | 671B |
| Family | Qwen | DeepSeek |
| Author | Alibaba | DeepSeek |
| Origin | CN | CN |
| License | Apache 2.0 | MIT |
| Context | 131,072 tokens | 128,000 tokens |
| Output | April 2025 | January 2025 |
Memory footprint
Approximate VRAM required for inference with an average context window. The winner (in green) is the model that uses less — smaller models have the edge.
| Quantization | Qwen 3 235B-A22B | DeepSeek R1 671B |
|---|---|---|
| Q4_K_M (lightweight) | 142 GB | 400 GB |
| Q5_K_M (balanced) | 170 GB | 480 GB |
| Q8 (near-lossless) | 250 GB | 720 GB |
| FP16 (max quality) | 470 GB | 1342 GB |
| CPU-only RAM | 160 GB | 512 GB |
Estimated speed (tokens/second)
Estimates based on the best quantization that can fit on each GPU. Actual figures depend on the prompt, context, and engine (llama.cpp, vLLM, MLX). Methodology.
| Reference GPU | Qwen 3 235B-A22B | DeepSeek R1 671B |
|---|---|---|
| RTX 4090 (24 GB) | ✗ too heavy | ✗ too heavy |
| RTX 4080 (16 GB) | ✗ too heavy | ✗ too heavy |
| RTX 3060 12GB (12 GB) | ✗ too heavy | ✗ too heavy |
| Apple M4 Pro (48 GB) (36 GB) | ✗ too heavy | ✗ too heavy |
Verdict by use case
For each common use case, we indicate which of the two is better suited based on its tags, size, and specialization.
Your private, free ChatGPT on your machine in 1 hour — LM Studio, Ollama, Open WebUI, your documents, no cloud.
- Lifetime online access
- PDF + files
- 30-day money-back guarantee
Strengths and weaknesses
Qwen 3 235B-A22B
235B/22B active MoE (128 experts, 8 active). AIME 2024 85.7, LiveCodeBench 70.7.
- Best open MoE at the time
- Active 22B → fast for its size
- AIME 2024 85.7
- At least 142 GB in Q4
- Multi-GPU deployment or Mac Studio required
Installation
DeepSeek R1 671B
MoE 671B/37B MIT. The open reference reasoning model. MATH-500 97.3, AIME 79.8, MMLU 90.8.
- MIT License
- Reference open reasoner
- MATH-500 97.3
- 400+ GB in Q4 — beyond the reach of a standard laptop/workstation
- For serious workstations or servers only
Installation
Frequently asked questions
Which model runs best on RTX 4090 (24 GB): Qwen 3 235B-A22B or DeepSeek R1 671B?
On a RTX 4090, Qwen 3 235B-A22B does not fit in 24 GB — part of it must be offloaded to the CPU; DeepSeek R1 671B does not fit in 24 GB either. See the configurator to test your exact GPU.
Which uses the least VRAM?
In Q4_K_M, Qwen 3 235B-A22B fits in 142 GB versus 400 GB for its rival—a gap of 258 GB, significant if you're targeting an RTX 3060 or a 4060 Ti 8 GB.
Can these models be used in commercial production?
Qwen 3 235B-A22B is under Apache 2.0 — free for commercial use. DeepSeek R1 671B is also free under the MIT license. For SaaS, favor Apache 2.0 or MIT.
Which one should you choose in 2026?
Depends on your primary constraint. Smaller / faster: Qwen 3 235B-A22B (235B). More capable: DeepSeek R1 671B (671B). If you're unsure, run the configurator with your GPU and use case — it will decide based on both.
Alternatives to consider
If neither works for you, here are some neighboring models you could explore.