Qwen2.5 72B Instruct
Alibaba Qwen2.5 72B Instruct is an instruction-tuned model with 128K context, long-form generation, structured data analysis, and multilingual instruction.
Model Overview
Capabilities, design details, and architectural traits
Qwen2.5 72B Instruct - Flagship Open-weight Instruction Model
Qwen2.5 72B Instruct is the flagship instruction-tuned model in the Qwen2.5 language model family. Its post-training focuses on improving long text generation, structured data analysis, and instruction following, while retaining broad multilingual capabilities across general-purpose tasks.
| Defining characteristic | Description |
|---|---|
| 128K context window | Processes long documents and extended conversations within a single context. |
| Long-form generation | Optimized to produce coherent responses over extended outputs. |
| Structured data analysis | Specifically improved for understanding and reasoning over structured information such as tables. |
| Instruction-tuned alignment | Uses supervised fine-tuning and multi-stage reinforcement learning to better follow user instructions and preferences. |
| Multilingual support | Supports more than 29 languages within the same instruction-tuned model. |
Qwen2.5 Post-training Focus
The model's identity comes from its extensive post-training pipeline. More than one million supervised fine-tuning samples are combined with multi-stage reinforcement learning to improve instruction following, human preference alignment, long-context behavior, and structured data understanding. These refinements distinguish it from the earlier Qwen2 instruction models.
Foundation for the Qwen2.5 Ecosystem
Qwen2.5 72B Instruct also serves as the foundation for several specialized successors, including Qwen2.5-Math, Qwen2.5-Coder, QwQ, and later multimodal Qwen2.5 models, making it the central general-purpose instruction model in the Qwen2.5 family.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Alibaba | Anthropic | Anthropic |
| Release Date | September 19, 2024 | September 22, 2026 | September 28, 2026 |
| Knowledge Cutoff | Jun 2024 | - | Jun 2026 |
| Context & Limits | |||
| Context Window | 131K | 1M Best Context Window | 1M Best Context Window |
| Pricing (per 1M tokens) | |||
| Input Pricing | $0.47 Best Input Pricing | $4 | $2 |
| Output Pricing | $0.50 Best Output Pricing | $20 | $10 |
| Modalities | |||
| Inputs | text | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 7.7 | 57.6 Best Intelligence Index | 56.0 |
| Coding Index | - | - | - |
| Agentic Index | - | - | - |
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
SciCode Benchmark
Scientific coding and mathematical modeling.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Explore more from Alibaba
Other models by Alibaba
Qwen3.8 Max
Qwen3.8 2.4T A95B
Qwen3.8-Flash-Next
Qwen3.8 27B
Top AI Models
Leading alternatives by intelligence score
