Qwen2.5 32B Instruct

Qwen

Qwen2.5-32B-Instruct is a 32 billion parameter language model from the Qwen2.5 series, fine-tuned for instruction following. This model excels at producing extended content exceeding 8,000 tokens, interpreting structured data formats like tables, and generating structured outputs, with a focus on JSON. It also boasts multilingual support for more than 29 languages.

Model Specifications

Technical details and capabilities of Qwen2.5 32B Instruct

Core Specifications

32.5B Parameters

Model size and complexity

18000.0B Training Tokens

Amount of data used in training

131.1K / 8.2K

Input / Output tokens

September 18, 2024

Release date

Capabilities & License

Multimodal Support

Not Supported

Web Hydrated

No

License

apache-2.0

Resources

API Reference

https://www.alibabacloud.com/help/en/model-studio/developer-reference/use-qwen-by-calling-api

Code Repository

https://github.com/QwenLM/Qwen2.5

Performance Insights

Check out how Qwen2.5 32B Instruct handles various AI tasks through comprehensive benchmark results.

100

75

50

25

0

95.9

GSM8K

95.9

(96%)

88.4

HumanEval

88.4

(88%)

85.2

HellaSwag

85.2

(85%)

84.5

BBH

84.5

(85%)

84

MBPP

84

(84%)

83.9

MMLU-Redux

83.9

(84%)

83.3

MMLU

83.3

(83%)

83.1

MATH

83.1

(83%)

82

Winogrande

82

(82%)

80.9

MMLU-STEM

80.9

(81%)

75.4

MultiPL-E

75.4

(75%)

70.4

ARC-C

70.4

(70%)

69

MMLU-Pro

69

(69%)

67.2

MBPP+

67.2

(67%)

57.8

TruthfulQA

57.8

(58%)

52.4

HumanEval+

52.4

(52%)

49.5

GPQA

49.5

(50%)

44.1

TheoremQA

44.1

(44%)

GSM8K

HumanEval

HellaSwag

BBH

MBPP

MMLU-Redux

MMLU

MATH

Winogrande

MMLU-STEM

MultiPL-E

ARC-C

MMLU-Pro

MBPP+

TruthfulQA

HumanEval+

GPQA

TheoremQA

Model Comparison

See how Qwen2.5 32B Instruct stacks up against other leading models across key performance metrics.

100

80

60

40

20

0

83.3

MMLU - Qwen2.5 32B Instruct

83.3

(83%)

86

MMLU - Llama 3.3 70B Instruct

86

(86%)

85.9

MMLU - Nova Pro

85.9

(86%)

79.7

MMLU - Qwen2.5 14B Instruct

79.7

(80%)

84.8

MMLU - Phi-4

84.8

(85%)

88.7

MMLU - GPT-4o

88.7

(89%)

49.5

GPQA - Qwen2.5 32B Instruct

49.5

(50%)

50.5

GPQA - Llama 3.3 70B Instruct

50.5

(51%)

46.9

GPQA - Nova Pro

46.9

(47%)

45.5

GPQA - Qwen2.5 14B Instruct

45.5

(46%)

56.1

GPQA - Phi-4

56.1

(56%)

53.6

GPQA - GPT-4o

53.6

(54%)

83.1

MATH - Qwen2.5 32B Instruct

83.1

(83%)

77

MATH - Llama 3.3 70B Instruct

77

(77%)

76.6

MATH - Nova Pro

76.6

(77%)

80

MATH - Qwen2.5 14B Instruct

80

(80%)

80.4

MATH - Phi-4

80.4

(80%)

76.6

MATH - GPT-4o

76.6

(77%)

88.4

HumanEval - Qwen2.5 32B Instruct

88.4

(88%)

88.4

HumanEval - Llama 3.3 70B Instruct

88.4

(88%)

89

HumanEval - Nova Pro

89

(89%)

83.5

HumanEval - Qwen2.5 14B Instruct

83.5

(84%)

82.6

HumanEval - Phi-4

82.6

(83%)

90.2

HumanEval - GPT-4o

90.2

(90%)

MMLU

GPQA

MATH

HumanEval

Qwen2.5 32B Instruct

Llama 3.3 70B Instruct

Nova Pro

Qwen2.5 14B Instruct

Phi-4

GPT-4o

Detailed Benchmarks

Dive deeper into Qwen2.5 32B Instruct's performance across specific task categories. Expand each section to see detailed metrics and comparisons.

Math

GSM8K

97.1%

Llama 3.1 405B Instruct

96.8%

Claude 3.5 Sonnet

96.4%

Claude 3.5 Sonnet

96.4%

Qwen2.5 32B Instruct

95.9%

95.9%

Qwen2.5 72B Instruct

95.8%

95.1%

68.6%

Current model

Other models

Avg (93.1%)

Coding

HumanEval

Claude 3.5 Sonnet

93.7%

89.0%

Llama 3.1 405B Instruct

89.0%

Mistral Small 3.1 24B

88.4%

Llama 3.3 70B Instruct

88.4%

Qwen2.5 32B Instruct

88.4%

Qwen2.5-Coder 7B Instruct

88.4%

88.4%

Claude 3.5 Haiku

88.1%

Ministral 8B Instruct

34.8%

Current model

Other models

Avg (83.7%)

HumanEval+

82.8%

Qwen2.5 32B Instruct

52.4%

Qwen2.5 14B Instruct

51.2%

Current model

Other models

Avg (62.1%)

MBPP

Qwen2.5-Coder 32B Instruct

90.2%

Qwen2.5 72B Instruct

88.2%

Qwen2.5 32B Instruct

84.0%

Qwen2.5-Coder 7B Instruct

83.5%

Qwen2.5 14B Instruct

82.0%

Phi-3.5-MoE-instruct

80.8%

Qwen2 72B Instruct

80.2%

Qwen2.5 7B Instruct

79.2%

52.4%

Current model

Other models

Avg (80.1%)

MBPP+

Qwen2.5 32B Instruct

67.2%

Qwen2.5 14B Instruct

63.2%

Current model

Other models

Avg (65.2%)

Reasoning

HellaSwag

95.4%

88.6%

Qwen2 72B Instruct

87.6%

86.4%

85.9%

Qwen2.5 32B Instruct

85.2%

Phi-3.5-MoE-instruct

83.8%

Mistral NeMo Instruct

83.5%

Qwen2.5-Coder 32B Instruct

83.0%

Phi-3.5-mini-instruct

69.4%

Current model

Other models

Avg (84.9%)

Knowledge

MMLU

91.8%

84.8%

Mistral Large 2

84.0%

84.0%

Llama 3.1 70B Instruct

83.6%

Qwen2.5 32B Instruct

83.3%

Qwen2 72B Instruct

82.3%

82.0%

81.3%

Llama 3.2 3B Instruct

63.4%

Current model

Other models

Avg (82.0%)

GPQA

87.7%

51.0%

Llama 3.1 405B Instruct

50.7%

Llama 3.3 70B Instruct

50.5%

50.4%

Qwen2.5 32B Instruct

49.5%

Qwen2.5 72B Instruct

49.0%

48.0%

46.9%

Qwen2 7B Instruct

25.3%

Current model

Other models

Avg (50.9%)

MATH

97.9%

Gemini 2.0 Flash

89.7%

89.0%

86.5%

85.5%

Qwen2.5 32B Instruct

83.1%

Qwen2.5 72B Instruct

83.1%

80.4%

Qwen2.5 14B Instruct

80.0%

32.6%

Current model

Other models

Avg (80.8%)

Hallucination

TruthfulQA

Phi-3.5-MoE-instruct

77.5%

Phi-3.5-mini-instruct

64.0%

Llama 3.1 Nemotron 70B Instruct

58.6%

Qwen2.5 14B Instruct

58.4%

Jamba 1.5 Large

58.3%

Qwen2.5 32B Instruct

57.8%

Qwen2 72B Instruct

54.8%

Qwen2.5-Coder 32B Instruct

54.2%

54.1%

Mistral NeMo Instruct

50.3%

Current model

Other models

Avg (58.8%)

Non categorized

MMLU-Pro

84.0%

72.6%

72.0%

Qwen2.5 72B Instruct

71.1%

70.4%

Qwen2.5 32B Instruct

69.0%

Llama 3.3 70B Instruct

68.9%

68.5%

Gemini 1.5 Flash

67.3%

Qwen2.5-Coder 7B Instruct

40.1%

Current model

Other models

Avg (68.4%)

MMLU-Redux

92.9%

89.1%

Qwen2.5 72B Instruct

86.8%

Qwen2.5 32B Instruct

83.9%

Qwen2.5 14B Instruct

80.0%

Qwen2.5-Coder 32B Instruct

77.5%

Qwen2.5 7B Instruct

75.4%

Qwen2.5-Coder 7B Instruct

66.6%

Current model

Other models

Avg (81.5%)

BBH

87.6%

86.9%

Qwen2.5 32B Instruct

84.5%

84.3%

82.4%

Qwen2 72B Instruct

82.4%

79.5%

Qwen2.5 14B Instruct

78.2%

70.6%

Current model

Other models

Avg (81.8%)

ARC-C

94.8%

92.4%

90.2%

Llama 3.2 3B Instruct

78.6%

Qwen2.5 32B Instruct

70.4%

Qwen2.5 14B Instruct

67.3%

Current model

Other models

Avg (82.3%)

Winogrande

87.5%

85.4%

Qwen2 72B Instruct

85.1%

Llama 3.1 Nemotron 70B Instruct

84.5%

Qwen2.5 32B Instruct

82.0%

80.6%

Mistral NeMo Instruct

76.8%

Ministral 8B Instruct

75.3%

Current model

Other models

Avg (82.2%)

TheoremQA

Qwen2.5 32B Instruct

44.1%

Qwen2.5-Coder 32B Instruct

43.1%

Qwen2.5 14B Instruct

43.0%

Qwen2.5-Coder 7B Instruct

34.0%

Current model

Other models

Avg (41.1%)

MMLU-STEM

Qwen2.5 32B Instruct

80.9%

Qwen2.5 14B Instruct

76.4%

Current model

Other models

Avg (78.6%)

MultiPL-E

Qwen2.5 32B Instruct

75.4%

Qwen2.5 72B Instruct

75.1%

Qwen2.5 14B Instruct

72.8%

Qwen2.5 7B Instruct

70.4%

Qwen2 72B Instruct

69.2%

Qwen2 7B Instruct

59.1%

Current model

Other models

Avg (70.3%)

Providers Pricing Coming Soon

We're working on gathering comprehensive pricing data from all major providers for Qwen2.5 32B Instruct. Compare costs across platforms to find the best pricing for your use case.

OpenAI

Anthropic

Google

Mistral AI

Cohere

Share your feedback

Hi, I'm Charlie Palars, the founder of Deepranking.ai. I'm always looking for ways to improve the site and make it more useful for you. You can write me through this form or directly through X at @palarsio.

Your feedback helps us improve our service

Stay Ahead with AI Updates

Get insights on Gemini Pro 2.5, Sonnet 3.7 and more top AI models