Mixtral 8x22B vs Qwen 2.5 72B
Architecture Comparison
SpecMixtral 8x22BQwen 2.5 72B
TypeMOEDENSE
Total Parameters141B72.7B
Active Parameters39B72.7B
Layers5680
Hidden Dimension6,1448,192
Attention Heads4864
KV Heads88
Context Length65,536131,072
Precision (default)BF16BF16
Total Experts8N/A
Active Experts2N/A
Memory Requirements
PrecisionMixtral 8x22BQwen 2.5 72B
BF16 Weights282.0 GB145.4 GB
FP8 Weights141.0 GB72.7 GB
INT4 Weights70.5 GB36.4 GB
KV-Cache / Token229376 B327680 B
Activation Estimate2.50 GB2.50 GB
Minimum GPUs Needed (BF16)
H100 SXM5 GPUs3 GPUs
L40S7 GPUs4 GPUs
Quality Benchmarks
BenchmarkMixtral 8x22BQwen 2.5 72B
Overall7384
MMLU77.885.3
HumanEval46.056.0
GSM8K78.491.6
MT-Bench80.086.0
Mixtral 8x22B
MMLU
77.8
HumanEval
46.0
GSM8K
78.4
MT-Bench
80.0
Qwen 2.5 72B
MMLU
85.3
HumanEval
56.0
GSM8K
91.6
MT-Bench
86.0
Capabilities
FeatureMixtral 8x22BQwen 2.5 72B
Tool Use✓ Yes✓ Yes
Vision✗ No✗ No
Code✓ Yes✓ Yes
Math✓ Yes✓ Yes
Reasoning✗ No✗ No
Multilingual✓ Yes✓ Yes
Structured Output✓ Yes✓ Yes
API Pricing Comparison
Cheapest Output (Mixtral 8x22B)
$1.20/M
Input: $1.20/M
Cheapest Output (Qwen 2.5 72B)
$0.90/M
Input: $0.90/M
| Provider | Mixtral 8x22B In $/M | Out $/M | Qwen 2.5 72B In $/M | Out $/M |
|---|---|---|---|---|
| together | $1.20 | $1.20 | $0.90 | $0.90 |
| fireworks | — | — | $0.90 | $0.90 |
| mistral | $2.00 | $6.00 | — | — |
Recommendation Summary
- ‣Qwen 2.5 72B scores higher on overall quality (84 vs 73).
- ‣Qwen 2.5 72B is cheaper per output token ($0.90/M vs $1.20/M).
- ‣Qwen 2.5 72B has a smaller memory footprint (145.4 GB vs 282.0 GB BF16), making it easier to deploy on fewer GPUs.
- ‣Qwen 2.5 72B supports a longer context window (131,072 vs 65,536 tokens).
- ‣Mixtral 8x22B uses MOE architecture while Qwen 2.5 72B uses DENSE. MoE models activate fewer parameters per token, improving inference efficiency.
- ‣Qwen 2.5 72B is stronger at code generation (HumanEval: 56.0 vs 46.0).
- ‣Qwen 2.5 72B is better at math reasoning (GSM8K: 91.6 vs 78.4).