Qwen 3 32B vs Mistral Small 24B
Architecture Comparison
SpecQwen 3 32BMistral Small 24B
TypeDENSEDENSE
Total Parameters32.8B24B
Active Parameters32.8B24B
Layers6440
Hidden Dimension5,1206,144
Attention Heads4048
KV Heads88
Context Length131,07232,768
Precision (default)BF16BF16
Memory Requirements
PrecisionQwen 3 32BMistral Small 24B
BF16 Weights65.6 GB48.0 GB
FP8 Weights32.8 GB24.0 GB
INT4 Weights16.4 GB12.0 GB
KV-Cache / Token262144 B163840 B
Activation Estimate2.00 GB1.50 GB
Minimum GPUs Needed (BF16)
H100 SXM1 GPU1 GPU
L40S2 GPUs2 GPUs
Quality Benchmarks
BenchmarkQwen 3 32BMistral Small 24B
Overall8068
MMLU82.072.0
HumanEval55.045.0
GSM8K90.070.0
MT-Bench84.077.0
Qwen 3 32B
MMLU
82.0
HumanEval
55.0
GSM8K
90.0
MT-Bench
84.0
Mistral Small 24B
MMLU
72.0
HumanEval
45.0
GSM8K
70.0
MT-Bench
77.0
Capabilities
FeatureQwen 3 32BMistral Small 24B
Tool Use✓ Yes✓ Yes
Vision✗ No✗ No
Code✓ Yes✓ Yes
Math✓ Yes✓ Yes
Reasoning✓ Yes✗ No
Multilingual✓ Yes✓ Yes
Structured Output✓ Yes✓ Yes
API Pricing Comparison
Cheapest Output (Qwen 3 32B)
$0.80/M
Input: $0.80/M
Cheapest Output (Mistral Small 24B)
$0.30/M
Input: $0.10/M
| Provider | Qwen 3 32B In $/M | Out $/M | Mistral Small 24B In $/M | Out $/M |
|---|---|---|---|---|
| together | $0.80 | $0.80 | $0.30 | $0.30 |
| mistral | — | — | $0.10 | $0.30 |
| fireworks | $0.90 | $0.90 | — | — |
Recommendation Summary
- ‣Qwen 3 32B scores higher on overall quality (80 vs 68).
- ‣Mistral Small 24B is cheaper per output token ($0.30/M vs $0.80/M).
- ‣Mistral Small 24B has a smaller memory footprint (48.0 GB vs 65.6 GB BF16), making it easier to deploy on fewer GPUs.
- ‣Qwen 3 32B supports a longer context window (131,072 vs 32,768 tokens).
- ‣Qwen 3 32B is stronger at code generation (HumanEval: 55.0 vs 45.0).
- ‣Qwen 3 32B is better at math reasoning (GSM8K: 90.0 vs 70.0).