Qwen2.5-72B-Instruct
- In Input modalities: Text Out Output modalities: Text
- Tool use
- Structured output
- Open weights
Prices
1 offering · prices as listedqwen2.5-72b-instruct| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard | ¥4.00 | ¥12.00 |
Prices from 怎么查询模型训练价格和模型部署价格-大模型服务平台百炼(Model Studio)-阿里云帮助中心 (opens in a new tab) (checked Sep 29, 2026)
Standard: the price a request with no conditions pays (the headline price elsewhere on the site).
Details
- Reasoning
- —
- Knowledge cutoff
- Apr 2024
- Family
- qwen
- Model ID
Not stated on Alibaba Qwen’s pages: max input, reasoning efforts, reasoning can be off, max thinking budget, built-in tools, licence, aliases.
Run locally
Weights: Qwen/Qwen2.5-72B-Instruct- Parameters 72.7B
- Dense
- Native precision BF16
- Layers 80
- KV cache per token 320 KB
- Config positions 32K (config.json max_position_embeddings)
- API context 977K (maker’s API)
- Needs
- ≈ 40.3 GB (weights 36.0 + KV cache 2.5 + overhead 1.8)
- Usable
- 21.6 GB (90% of its memory)
- In system memory
- 18.9 GB · 38/80 layers on the GPU (--n-gpu-layers 38)
- Generation speed
- 2–3 tok/s
- First token
- about 2.35–3.52 s for a 2K prompt
| Precision | Weights | Needs at 8K | On your hardware | Longest context |
|---|---|---|---|---|
| BF16 (native)maker’s file | 135.4 GB | ≈ 144.7 | ✕Doesn’t fit | — |
| Q8_0maker’s file | 72.2 GB | ≈ 78.3 | ✕Doesn’t fit | — |
| INT8 (GPTQ)maker’s file | 71.8 GB | ≈ 77.9 | ✕Doesn’t fit | — |
| FP8 | 67.7 GB | ≈ 73.6 | ✕Doesn’t fit | — |
| Q6_Kmaker’s file | 55.7 GB | ≈ 61.0 | ✕Doesn’t fit | — |
| Q5_K_Mmaker’s file | 48.1 GB | ≈ 53.0 | ✕Doesn’t fit | — |
| Q5_0maker’s file | 46.9 GB | ≈ 51.7 | ✕Doesn’t fit | — |
| Q4_K_Mmaker’s file | 41.0 GB | ≈ 45.5 | ✕Doesn’t fit | — |
| Q4_0maker’s file | 38.5 GB | ≈ 43.0 | ○Needs system memory 2–3 tok/s | — |
| INT4 (AWQ / GPTQ)maker’s file | 38.7 GB | ≈ 43.2 | ○Needs system memory 2–3 tok/s | — |
| IQ4_XS | 36.0 GB | ≈ 40.3 | ○Needs system memory 2–3 tok/s | — |
| Q3_K_Mmaker’s filequality drops | 33.0 GB | ≈ 37.2 | ○Needs system memory 2–4 tok/s | — |
| IQ3_XXSquality drops | 25.9 GB | ≈ 29.7 | ○Needs system memory 4–6 tok/s | — |
| Q2_Kmaker’s filequality drops | 25.5 GB | ≈ 29.2 | ○Needs system memory 5–7 tok/s | — |
| IQ2_XXSquality drops | 17.4 GB | ≈ 20.8 | ●Runs well 24–35 tok/s | 10K |
Click a row to choose a precision. “Maker’s file” sizes are the maker’s own files; others are parameters × bits per weight.
Compared with other hardware
Each at the highest precision it runs. Click one for details.
Smallest that runs it (Q4_K_M, 8K): GPU NVIDIA H100 SXM; unified memory Apple M5 Max (40-core GPU) · 64 GB.
Run commands
The maker publishes no files at this precision. Choose its native precision or one of its own builds, or quantise it yourself with llama.cpp.
Memory to fine-tune
Batch 1, 2,048-token sequences, gradient checkpointing; LoRA and QLoRA train rank-16 adapters on the attention projections.
Estimates; facts from the maker’s Hugging Face repositories, as of 2026-11-06. How it’s estimated
Cited scores
From third-party leaderboards, not measured by us- LMArena Text 1,303
#147 of 163 listed models · CI 1,299–1,307 · 39,406 votes
- Epoch AI · GPQA Diamond 49.1%
#127 of 168 listed models · ±2.69
- Epoch AI · MATH Level 5 63.2%
#30 of 64 listed models · ±1.07
- Epoch AI · OTIS Mock AIME 2024–2025 8.1%
#132 of 159 listed models · ±2.61
- LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · License · as of Oct 28, 2026
- Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · License · as of Oct 18 – Nov 4, 2026
- changes we made
Changes
All changes →No changes recorded yet.
Sources
- Text generation models - Qwen, DeepSeek, GLM | Model Studio - Alibaba Cloud Model Studio - Alibaba Cloud Documentation Center (opens in a new tab) docs.modelstudio.console.alibabacloud.com · checked Oct 15, 2026
- Qwen/Qwen2.5-72B-Instruct README.md (opens in a new tab) huggingface.co · checked Nov 1, 2026
- 怎么查询模型训练价格和模型部署价格-大模型服务平台百炼(Model Studio)-阿里云帮助中心in Chinese (opens in a new tab) help.aliyun.com · checked Sep 29, 2026
- models.dev: qwen2-5-72b-instruct (community data, MIT) (opens in a new tab) github.com · checked Sep 30, 2026
- Qwen2.5: A Party of Foundation Models! | Qwen (opens in a new tab) qwenlm.github.io · checked Sep 30, 2026
- Qwen/Qwen2.5-72B-Instruct model card (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen2.5-72B-Instruct config.json (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen2.5-72B-Instruct files (opens in a new tab) huggingface.co · checked Nov 3, 2026
- Qwen/Qwen2.5-72B-Instruct-AWQ files (opens in a new tab) huggingface.co · checked Nov 2, 2026
- Qwen/Qwen2.5-72B-Instruct-GGUF files (opens in a new tab) huggingface.co · checked Nov 3, 2026
- Qwen/Qwen2.5-72B-Instruct-GPTQ-Int4 files (opens in a new tab) huggingface.co · checked Nov 1, 2026
- Qwen/Qwen2.5-72B-Instruct-GPTQ-Int8 files (opens in a new tab) huggingface.co · checked Oct 31, 2026
Alibaba Qwen’s official policies: data use, privacy, terms →
Some values on this page are still waiting to be checked against official pages: a filled dot is verified, a hollow one is not yet (with or without a source on file). How we verify data
Alibaba Qwen · China · qwen.ai
Related models
Snapshots, the same family and other models from this provider