Ministral 14B
Trinity Nano Preview
- Context
- 131Ktokens
- Released
- 2y ago
- Size
- 7Bparameters
Qwen2.5 7B Instruct is an open-weight, instruction-tuned language model from Alibaba and the successor to Qwen2 in the Qwen family. Announcing it as what the maker calls possibly the largest open-source release to date, Alibaba says the Qwen2.5 models were pretrained on up to 18 trillion tokens and improve on Qwen2 in knowledge, coding, mathematics, instruction following, and handling structured data. The model is built for general-purpose text work, including multilingual chat, role-play, and producing structured outputs such as JSON.
How it compares
Size class
Small7BAn open-weight model of 5 to 25 billion parameters, the size that runs on a single consumer GPU.
Open models of a similar size from other labs
Vs. your stack
Add the models you use to see how this one compares.
Sign inAdditional source evidence
- Source model
- alibaba/qwen2.5-7b-instruct
- Source family
- qwen
- Last updated
- 2024-09-19
- Model context
- 131K
- Model output limit
- 8K
- Open weights
- Yes
- File attachments
- No
- Reasoning
- No
- Tool calling
- Yes
- Temperature control
- Yes
Downloadable weights
Complete source record (JSON)
{
"id": "alibaba/qwen2.5-7b-instruct",
"name": "Qwen2.5 7B Instruct",
"description": "Small open Qwen2.5 instruction model for chat, structured output, and tool use",
"family": "qwen",
"last_updated": "2024-09-19",
"attachment": false,
"reasoning": false,
"tool_call": true,
"temperature": true,
"modalities": {
"input": [
"text"
],
"output": [
"text"
]
},
"limit": {
"context": 131072,
"output": 8192
},
"release_date": "2024-09-19",
"open_weights": true,
"license": "Apache 2.0",
"weights": [
{
"label": "Hugging Face",
"url": "https://huggingface.co/Qwen/Qwen2.5-7B-Instruct"
}
]
}Frequently asked questions
Running locally
Qwen2.5 7B Instruct needs an estimated 5.4 GB of VRAM at Q4, 9.1 GB at Q8 and 16.6 GB at FP16. These figures are estimates from the 7B parameter count, covering weights plus about 15% and 0.5 GB for a short context.
Yes, it fits on all four. A 24 GB GPU fits FP16 (estimated 16.6 GB), 16 GB and 12 GB GPUs fit Q8 (estimated 9.1 GB), and an 8 GB GPU fits Q4 (estimated 5.4 GB).
Yes. At Q4 (estimated 5.4 GB) it fits on an 8 GB Mac, the smallest common memory size, assuming macOS lets the GPU use about 75% of unified memory.
Privacy and data
Yes, the model has open weights, so you can self-host it for full control of your data.
Capabilities and specs
The context window is 131K tokens (131,072) and the maximum output is 8K tokens (8,192).
No. The only supported input modality is text, so it cannot read images, audio or video.
Yes, tool calling (function calling) is supported.
No, a reasoning or thinking mode is not supported.
Quality and benchmarks
The closest alternatives, matched on model kind and release date with no comparable price, are Llama-3.2-1B by Meta, Llama-3.2-3B by Meta and Molmo 72B by Ai2.
Llama-3.2-1BQwen2.5 7B Instruct vs Llama-3.2-1BLlama-3.2-3BQwen2.5 7B Instruct vs Llama-3.2-3BMolmo 72BQwen2.5 7B Instruct vs Molmo 72B
Versions and lineage
Qwen2.5 7B Instruct has a 131K token context (131,072) and was released September 19, 2024, while Qwen2.5-Coder-0.5B has a 33K token context (32,768) and was released later, on November 12, 2024.
Licensing and access
It is open weights, released under the Apache 2.0 license.
Yes, the Apache 2.0 license is permissive and allows commercial use.
Basics
Qwen made the model, and it was released on September 19, 2024.
Yes, it has open weights, so the model can be downloaded and used for free.
Generated from Vertical Tabs data, updated November 10, 2026.
Discussion
No comments yet
Be the first to share your thoughts.