Qwen 3 235B-A22B vs DeepSeek R1 671B
Side-by-side specs, benchmarks, and a verdict by use case.
Updated 2026-07-13
| Spec | Qwen 3 235B-A22B | DeepSeek R1 671B |
|---|---|---|
| Parameters | 235B | 671B |
| Author | Alibaba | DeepSeek |
| License | Apache 2.0 | MIT |
| Context window | 0k | 0k |
| VRAM at Q4 | 142 GB | 400 GB |
| VRAM at Q5 | 170 GB | 480 GB |
| VRAM at Q8 | 250 GB | 720 GB |
| VRAM at FP16 | 470 GB | 1342 GB |
| Use cases | chat, general, reasoning, multilingual, moe | reasoning, moe |
Verdict
DeepSeek R1 671B is significantly larger (671B vs 235B), so expect higher quality but heavier VRAM and slower throughput.
The two models at a glance
About Qwen 3 235B-A22B
Alibaba's flagship MoE — 235B total, 22B active per token across 128 experts. Hits 85.7 on AIME 2024 and 70.7 on LiveCodeBench, putting it in frontier-open territory. Strengths: Frontier-open scores on AIME 2024 (85.7) and LiveCodeBench (70.7), Only 22B active parameters — fast for its total size, Instruct-2507 and Thinking-2507 variants available, Apache 2.0.
About DeepSeek R1 671B
The reference open reasoning model — a 671B MoE with 37B active, released under MIT. Scores 97.3 on MATH-500, 79.8 on AIME, and 90.8 on MMLU. Strengths: MIT license — no commercial restrictions, Reference open reasoning model, MATH-500 score of 97.3, R1-0528 update further sharpens reasoning.
How they compare
Qwen 3 235B-A22B comes from Alibaba and DeepSeek R1 671B from DeepSeek, they belong to the Qwen and DeepSeek families respectively. This comparison is built entirely from structured specs — parameter count, VRAM by quantization, context window, license, and published benchmark scores — so the verdict below reflects measurable differences rather than marketing claims.
At 235B vs 671B parameters, DeepSeek R1 671B is the larger of the two. At Q4, Qwen 3 235B-A22B fits in about 142 GB of VRAM versus 400 GB for the other — a 258 GB difference that matters on consumer GPUs.
The two models target different sweet spots: Qwen 3 235B-A22B is tuned for chat, general, reasoning, multilingual, moe, while DeepSeek R1 671B leans toward reasoning, moe. Match the model to your dominant workload rather than to raw size.
On a typical mid-range GPU, Qwen 3 235B-A22B pushes roughly 12 tokens/sec versus 5, so it is the more responsive choice for interactive or high-volume use. For long-context work, Qwen 3 235B-A22B offers the bigger window (128k vs 125k tokens).
Memory, quantization & throughput
Across quantization levels, Qwen 3 235B-A22B requires Q4 ≈ 142 GB, Q5 ≈ 170 GB, Q8 ≈ 250 GB, FP16 ≈ 470 GB, while DeepSeek R1 671B requires Q4 ≈ 400 GB, Q5 ≈ 480 GB, Q8 ≈ 720 GB, FP16 ≈ 1342 GB. In practice Qwen 3 235B-A22B spills past 24 GB even at Q4, so plan your GPU around the Q4 or Q5 figure unless you specifically need the higher fidelity of Q8 or FP16.
Without a GPU, Qwen 3 235B-A22B needs roughly 160 GB of system RAM to run on CPU and DeepSeek R1 671B about 512 GB — workable for offline use but far slower than GPU inference. On a mid-range GPU you can expect on the order of 12 tokens/sec from Qwen 3 235B-A22B and 5 from DeepSeek R1 671B, scaling up to 28 and 15 tokens/sec on high-end hardware.
Benchmark scores
Reported benchmarks for Qwen 3 235B-A22B: AIME 2024 85.7, AIME 2025 81.5, LiveCodeBench v5 70.7.
Reported benchmarks for DeepSeek R1 671B: MMLU 90.8, GPQA Diamond 71.5, MATH-500 97.3.
Bottom line: which should you pick?
- Pick Qwen 3 235B-A22B for long-context work (up to 128k tokens).
- Pick Qwen 3 235B-A22B for lower VRAM and faster inference; pick DeepSeek R1 671B for maximum headline quality.
- Pick Qwen 3 235B-A22B if your workload is chat, general, multilingual.
Which GPU should you buy to run DeepSeek R1 671B?
To run DeepSeek R1 671B locally at Q4, you need ~400 GB of VRAM. The best value for this is a Apple Mac Studio (64+ GB unified memory).
As an Amazon Associate, BestLLMfor earns from qualifying purchases, at no extra cost to you. It does not influence our independent rankings.
Frequently asked questions
What is the difference between Qwen 3 235B-A22B and DeepSeek R1 671B?
The headline differences: Qwen 3 235B-A22B is a 235B model and DeepSeek R1 671B is 671B; their context windows differ (128k vs 125k tokens); they ship under different licenses (Apache 2.0 vs MIT). Below we break down VRAM by quantization, benchmark scores, and a use-case verdict so you can pick the right one.
Can Qwen 3 235B-A22B and DeepSeek R1 671B run on a 24 GB GPU?
At a Q4 quantization, Qwen 3 235B-A22B needs about 142 GB of VRAM and needs more than 24 GB (multi-GPU or heavier offload); DeepSeek R1 671B needs about 400 GB and needs more than 24 GB. Qwen 3 235B-A22B is the lighter option for tight VRAM budgets.
Which is faster, Qwen 3 235B-A22B or DeepSeek R1 671B?
Qwen 3 235B-A22B is the smaller model (235B vs 671B), so on the same hardware it runs faster and uses less memory. The larger model trades speed for headline quality.
What licenses do Qwen 3 235B-A22B and DeepSeek R1 671B use?
Qwen 3 235B-A22B is licensed under Apache 2.0 and DeepSeek R1 671B under MIT.
Which has the longer context window, Qwen 3 235B-A22B or DeepSeek R1 671B?
Qwen 3 235B-A22B has the larger context window (128k vs 125k tokens), so it handles longer documents and codebases in a single prompt.