Compare · ModelsLive · 3 picked · head to head
Qwen2.5 72B Instruct vs Qwen2.5 32B Instruct vs Qwen2-72B
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
Qwen2.5 72B Instruct wins on 10/15 benchmarks
Qwen2.5 72B Instruct wins 10 of 15 shared benchmarks. Leads in general · language · math.
Category leads
knowledge·Qwen2-72Bgeneral·Qwen2.5 72B Instructlanguage·Qwen2.5 72B Instructmath·Qwen2.5 72B Instructreasoning·Qwen2-72Bcoding·Qwen2.5 72B Instructagentic·Qwen2.5 72B Instruct
Hype vs Reality
Attention vs performance
Qwen2.5 72B Instruct
#128 by perf·#2 by attention
Qwen2.5 32B Instruct
#222 by perf·#2 by attention
Qwen2-72B
#203 by perf·#2 by attention
Best value
Qwen2.5 72B Instruct
Qwen2.5 72B Instruct
129.5 pts/$
$0.38/M
Qwen2.5 32B Instruct
n/a
no price
Qwen2-72B
n/a
no price
Vendor risk
Who is behind the model
Alibaba (Qwen)
$293.0B·Tier 1
Alibaba (Qwen)
$293.0B·Tier 1
Alibaba (Qwen)
$293.0B·Tier 1
Head to head
15 benchmarks · 3 models
Qwen2.5 72B InstructQwen2.5 32B InstructQwen2-72B
GPQA diamond
Qwen2.5 72B Instruct leads by +4.1
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Qwen2.5 72B Instruct
32.2
Qwen2.5 32B Instruct
28.1
Qwen2-72B
21.0
BBH (HuggingFace)
Qwen2.5 72B Instruct leads by +5.4
Qwen2.5 72B Instruct
61.9
Qwen2.5 32B Instruct
56.5
Qwen2-72B
51.9
GPQA
Qwen2-72B leads by +2.6
Qwen2.5 72B Instruct
16.7
Qwen2.5 32B Instruct
11.7
Qwen2-72B
19.2
IFEval
Qwen2.5 72B Instruct leads by +2.9
Qwen2.5 72B Instruct
86.4
Qwen2.5 32B Instruct
83.5
Qwen2-72B
38.2
MATH Level 5
Qwen2.5 32B Instruct leads by +2.7
Qwen2.5 72B Instruct
59.8
Qwen2.5 32B Instruct
62.5
Qwen2-72B
31.1
MMLU-PRO
Qwen2-72B leads by +0.7
Qwen2.5 72B Instruct
51.4
Qwen2.5 32B Instruct
51.9
Qwen2-72B
52.6
MUSR
Qwen2-72B leads by +6.2
Qwen2.5 72B Instruct
11.7
Qwen2.5 32B Instruct
13.5
Qwen2-72B
19.7
MATH level 5
Qwen2.5 72B Instruct leads by +7.1
MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics.
Qwen2.5 72B Instruct
63.2
Qwen2.5 32B Instruct
56.1
Qwen2-72B
39.1
Aider · Code Editing
Qwen2.5 72B Instruct leads by +9.8
Qwen2.5 72B Instruct
65.4
Qwen2-72B
55.6
CMMLU
Qwen2-72B leads by +4.0
Qwen2.5 72B Instruct
85.7
Qwen2-72B
89.7
Metr Time Horizons
Qwen2.5 72B Instruct leads by +5.9
Qwen2.5 72B Instruct
35.8
Qwen2-72B
29.9
MMLU
Qwen2.5 72B Instruct leads by +3.9
Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge.
Qwen2.5 72B Instruct
80.4
Qwen2-72B
76.5
OTIS Mock AIME 2024-2025
Qwen2.5 72B Instruct leads by +0.7
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Qwen2.5 72B Instruct
8.0
Qwen2.5 32B Instruct
7.3
The Agent Company
Qwen2.5 72B Instruct leads by +4.6
The Agent Company · tests AI agents on realistic corporate tasks like email management, code review, data analysis, and cross-tool workflows.
Qwen2.5 72B Instruct
5.7
Qwen2-72B
1.1
WeirdML
Qwen2.5 72B Instruct leads by +4.7
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Qwen2.5 72B Instruct
16.0
Qwen2-72B
11.3
Full benchmark table
| Benchmark | Qwen2.5 72B Instruct | Qwen2.5 32B Instruct | Qwen2-72B |
|---|---|---|---|
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 32.2 | 28.1 | 21.0 |
BBH (HuggingFace) | 61.9 | 56.5 | 51.9 |
GPQA | 16.7 | 11.7 | 19.2 |
IFEval | 86.4 | 83.5 | 38.2 |
MATH Level 5 | 59.8 | 62.5 | 31.1 |
MMLU-PRO | 51.4 | 51.9 | 52.6 |
MUSR | 11.7 | 13.5 | 19.7 |
MATH level 5 MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics. | 63.2 | 56.1 | 39.1 |
Aider · Code Editing | 65.4 | — | 55.6 |
CMMLU | 85.7 | — | 89.7 |
Metr Time Horizons | 35.8 | — | 29.9 |
MMLU Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge. | 80.4 | — | 76.5 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 8.0 | 7.3 | — |
The Agent Company The Agent Company · tests AI agents on realistic corporate tasks like email management, code review, data analysis, and cross-tool workflows. | 5.7 | — | 1.1 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 16.0 | — | 11.3 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $0.36 | $0.40 | 33K tokens (~16 books) | $3.70 | |
| — | — | — | — | |
| — | — | — | — |