Skip to content

Qwen 3 32B vs Phi-4

Alibaba
Qwen 3 32B

Alibaba · 32.8B params · Quality: 80

Microsoft
Phi-4

Microsoft · 14.7B params · Quality: 83

Architecture Comparison

SpecQwen 3 32BPhi-4
TypeDENSEDENSE
Total Parameters32.8B14.7B
Active Parameters32.8B14.7B
Layers6440
Hidden Dimension5,1205,120
Attention Heads4040
KV Heads810
Context Length131,07216,384
Precision (default)BF16BF16

Memory Requirements

PrecisionQwen 3 32BPhi-4
BF16 Weights65.6 GB29.4 GB
FP8 Weights32.8 GB14.7 GB
INT4 Weights16.4 GB7.3 GB
KV-Cache / Token262144 B204800 B
Activation Estimate2.00 GB1.50 GB

Minimum GPUs Needed (BF16)

H100 SXM1 GPU1 GPU
L40S2 GPUs1 GPU

Quality Benchmarks

BenchmarkQwen 3 32BPhi-4
Overall8083
MMLU82.084.8
HumanEval55.067.0
GSM8K90.093.0
MT-Bench84.085.0

Qwen 3 32B

MMLU
82.0
HumanEval
55.0
GSM8K
90.0
MT-Bench
84.0

Phi-4

MMLU
84.8
HumanEval
67.0
GSM8K
93.0
MT-Bench
85.0

Capabilities

FeatureQwen 3 32BPhi-4
Tool Use✓ Yes✓ Yes
Vision✗ No✗ No
Code✓ Yes✓ Yes
Math✓ Yes✓ Yes
Reasoning✓ Yes✓ Yes
Multilingual✓ Yes✓ Yes
Structured Output✓ Yes✓ Yes

API Pricing Comparison

Cheapest Output (Qwen 3 32B)

$0.80/M

Input: $0.80/M

Cheapest Output (Phi-4)

$0.14/M

Input: $0.07/M

ProviderQwen 3 32B In $/MOut $/MPhi-4 In $/MOut $/M
azure$0.07$0.14
together$0.80$0.80$0.20$0.20
fireworks$0.90$0.90

Recommendation Summary

  • Phi-4 scores higher on overall quality (83 vs 80).
  • Phi-4 is cheaper per output token ($0.14/M vs $0.80/M).
  • Phi-4 has a smaller memory footprint (29.4 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 16,384 tokens).
  • Phi-4 is stronger at code generation (HumanEval: 67.0 vs 55.0).
  • Phi-4 is better at math reasoning (GSM8K: 93.0 vs 90.0).

Compare Other Models