Versatile Qwen update with better language skills and wider support
100K+
Qwen2.5-7B-Instruct is an instruction-tuned large language model developed by Alibaba Cloud. It is part of the Qwen2.5 series, which includes models ranging from 0.5 to 72 billion parameters. This model offers significant improvements in knowledge, coding, and mathematical capabilities, along with enhanced instruction-following and long-text generation abilities. It supports a context length of up to 131,072 tokens and can generate outputs up to 8,192 tokens. Additionally, it provides multilingual support for over 29 languages, including Chinese, English, French, Spanish, Portuguese, German, Italian, Russian, Japanese, Korean, Vietnamese, Thai, and Arabic.
Qwen2.5-7B-Instruct is designed to assist in various natural language processing tasks, including:
| Attribute | Details |
|---|---|
| Provider | Alibaba Cloud |
| Architecture | qwen2 |
| Cutoff date | November 2024 (est) |
| Languages | Chinese, English, French, Spanish, Portuguese, German, Italian, Russian, Japanese, Korean, Vietnamese, Thai, Arabic, and more (29 languages) |
| Tool calling | ✅ |
| Input modalities | Text |
| Output modalities | Text |
| License | Apache 2.0 |
| Model variant | Parameters | Quantization | Context window | VRAM¹ | Size |
|---|---|---|---|---|---|
ai/qwen2.5:latestai/qwen2.5:7B-Q4_K_M | 7B | IQ2_XXS/Q4_K_M | 33K tokens | 4.83 GiB | 4.36 GB |
ai/qwen2.5:0.5B-F16 | 0.5B | F16 | 33K tokens | 1.38 GiB | 942.43 MB |
ai/qwen2.5:1.5B-F16 | 1.5B | F16 | 33K tokens | 3.39 GiB | 2.88 GB |
ai/qwen2.5:3B-Q4_K_M | 3B | IQ2_XXS/Q4_K_M | 33K tokens | 2.37 GiB | 1.79 GB |
ai/qwen2.5:3B-F16 | 3B | F16 | 33K tokens | 6.33 GiB | 5.75 GB |
ai/qwen2.5:7B-Q4_0 | 7B | Q4_0 | 33K tokens | 4.60 GiB | 4.12 GB |
ai/qwen2.5:7B-Q4_K_M | 7B | IQ2_XXS/Q4_K_M | 33K tokens | 4.83 GiB | 4.36 GB |
ai/qwen2.5:7B-F16 | 7B | F16 | 33K tokens | 13.93 GiB | 14.19 GB |
¹: VRAM estimated based on model characteristics.
latest→7B-Q4_K_M
First, pull the model:
docker model pull ai/qwen2.5
Then run the model:
docker model run ai/qwen2.5
For more information on Docker Model Runner, explore the documentation.
| Metrics | Benchmark | Qwen2.5-7B-Instruct |
|---|---|---|
| Knowledge & QA | MMLU-Pro | 56.3 |
| MMLU-redux | 75.4 | |
| GPQA | 36.4 | |
| Math & Reasoning | MATH | 75.5 |
| GSM8K | 91.6 | |
| Code | HumanEval | 84.8 |
| MBPP | 79.2 | |
| MultiPL-E | 70.4 | |
| LiveCodeBench 2305-2409 | 28.7 | |
| LiveBench 0831 | 35.9 | |
| Instruction Following | IFeval strict-prompt | 71.2 |
| Arena-Hard | 52.0 | |
| Alignment & Preference | AlignBench v1.1 | 7.33 |
| MTbench | 8.75 |
Content type
Model
Digest
sha256:b117fb2e0…
Size
4.1 GB
Last updated
12 months ago
docker model pull ai/qwen2.5:7B-Q4_0Pulls:
2,797
Last week