Qwen2-7B-Instruct

Qwen2 7B · Family: Qwen

Instruction-tuned 7B language model from Alibaba's Qwen2 series, open-weights under Apache 2.0.

✓ Active✓ Public access⚖ Open sourceLLMTool-using model📁 Qwen

Context window

128K

tokens

Parameters

7.07B

parameters

Release date

7 June 2024

🏢AlibabaProducer 🔬QwenResearch lab

Access:APIDownloadHostedDeployment:💻 Local☁ Cloud

Overview

Qwen2-7B-Instruct is an instruction-tuned large language model from the Qwen2 series, developed by the Qwen team at Alibaba Cloud and released on 7 June 2024. The base model has 7.07B parameters (5.98B non-embedding). It uses a Transformer architecture with SwiGLU activation, attention QKV bias and Grouped Query Attention (GQA), plus an improved tokenizer adaptive to multiple languages and code. The base model was pretrained on a 32,768-token context, while the instruction-tuned variant supports up to 131,072 tokens (128K) via the YaRN technique. Post-training combined supervised fine-tuning (SFT) and Direct Preference Optimization (DPO). The model is multilingual (English, Chinese and 27 additional languages) and released under the open Apache 2.0 license.

Classification

LLMTool-using model

Family: Qwen

Access & deployment

APIDownloadHosted

LocalCloud

Weights: Open source

Key parameters

📏 Context: 128K

🧩 Parameters: 7.07B

✓ Tools · ✓ Fine-tuning

📥 Input: text

Technical specification

Context window

128K

tokens

Parameters

7.07B

parameters

License

Apache 2.0

Hardware requirements

~15 GB VRAM in BF16/FP16; runs on a single 16 GB-class GPU (e.g. with quantization) or locally via llama.cpp/Ollama/vLLM.

Features:✓ Tool use✓ Fine-tuning

Modalities

⬇ Input

text

⬆ Output

textcode

Capabilities and applications

Native model capabilities

Language modeling

Ability to predict subsequent tokens and generate coherent natural-language text based on the preceding context.

Category: language

Instruction following

Precisely following instructions contained in the prompt: response format, length, style, constraints (e.g. 'reply in six words'). GPT-5.1 significantly improved this capability compared to GPT-5.

Category: language

Coding

Generating, analysing and modifying code in many programming languages. Covers writing functions, debugging, refactoring, code review, and creating tests. Measured by benchmarks such as HumanEval and SWE-bench.

Category: coding

Mathematical reasoning

The model's ability to solve mathematical tasks requiring multi-step reasoning — equations, proofs, combinatorics, geometry, calculus and competition-level problems.

Category: reasoning

Multilingual

Competence in many natural languages (from a few to over a hundred): understanding, generation, translation, and code-switching within a single conversation. Frontier models support a wide range of languages with comparable quality.

Category: language

Long context

Support for large context windows — tens to hundreds of thousands (or millions) of input tokens. Enables analysis of entire codebases, long documents, and many parallel conversations without losing earlier information. GPT-5.1 supports 400,000 tokens.

Category: language

Tool use

The model's ability to call external functions, APIs and tools during a conversation: calculator, search engine, code editor, database. The model decides when and how to use a tool and interprets its result.

Category: planning

Benchmark results

7 benchmarks

MMLU

accuracy

70.5