Skip to content
Weights Not Walls
Menu

Qwen

Qwen2.5-72B-Instruct

Reviewed
Size
72.7B parameterscounted from the safetensors header on the Hub
Licence
custom, read it, Qwen License AgreementOur reading, check the licence: A custom licence we have not read for this model: other makers use the same name for different texts. Read the model's own licence file before you use it.Licence text
On the Hub since
16 Sep 2024the date the repo was created
Context length
32k tokensfrom the model's config
Fine-tune
of Qwen/Qwen2.5-72BQwen2ForCausalLM
On Hugging Face
Qwen/Qwen2.5-72B-Instruct156k downloads in the last 30 days, 1.0k likes

Scores

11 independent results on 10 eval sets
GPQA diamondRe-run by Epoch AI
  • default setting49.1%±2.727 Jan 2025, log

Epoch's name: qwen2.5-72b-instruct. Best closed model on this set: gpt-6.2-astra_max, 96.8% (OpenAI). All 341 results

MATH level 5Re-run by Epoch AI
  • default setting63.2%±1.127 Jan 2025, log

Epoch's name: qwen2.5-72b-instruct. Best closed model on this set: gpt-5-2025-08-07_high, 98.1% (OpenAI). All 108 results

OTIS Mock AIME 2024-2025Re-run by Epoch AI
  • default setting8.1%±2.625 Feb 2025, log

Epoch's name: qwen2.5-72b-instruct. Best closed model on this set: claude-fable-5_high, 100.0% (Anthropic). All 318 results

BalrogReported

Epoch's name: qwen2.5-72b-instruct. Best closed model on this set: gpt-6-astra_max, 68.3% (OpenAI). All 41 results

DTBenchReported

Epoch's name: qwen2.5-72b-instruct. Best closed model on this set: claude-fable-5-2_unknown, 99.2% (Anthropic). All 241 results

LMCAReported

Epoch's name: qwen2.5-72b-instruct. Best closed model on this set: claude-fable-5-2_xhigh, 70.6% (Anthropic). All 203 results

Epoch's name: qwen2.5-72b-instruct. Best closed model on this set: claude-mythos-preview-early, 85.2% (Anthropic). All 49 results

MMLUReported

Epoch's name: qwen2.5-72b-instruct. Best closed model on this set: gpt-4o-2024-11-20, 88.1% (OpenAI). All 217 results

Epoch's name: qwen2.5-72b-instruct. Best closed model on this set: gemini-2.5-flash-preview-09-2025, 41.1% (Google DeepMind). All 16 results

WeirdMLReported

Epoch's name: qwen2.5-72b-instruct. Best closed model on this set: gpt-6-astra_promax, 93.6% (OpenAI). All 173 results

Claimed by the maker Claimed

The model card has no claimed scores in a machine-readable form (the card's model-index), so none are shown. The maker may give scores in the card text: read the card.

Other views, which we link to and do not copy: look this model up on Artificial Analysis and the Arena leaderboard.

Independent scores from Epoch AI (CC-BY 4.0), fetched 19 Nov 2026: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks’ [online resource]. Each row links its own source. What the badges mean.

Run it yourself

file sizes read from the Hub 19 Nov 2026

Memory needed = weights + context cache + 0.50 GB for the runtime. Context cache at 8k tokens: 2.5 GB, worked out from the model's config (Qwen/Qwen2.5-72B-Instruct), at 16-bit precision. Choose a machine above to see a verdict for each format. Method

GGUF for llama.cpp, Ollama, LM Studio and others

Qwen/Qwen2.5-72B-Instruct-GGUF (the maker's own conversion), 12k downloads. Sizes are the real files on the Hub, all parts added up.

FormatSizeMemory neededGet it
Q2_K25.5 GBfile size, 7 parts28.5 GB25.5 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf Qwen/Qwen2.5-72B-Instruct-GGUF:Q2_K
Download
hf download Qwen/Qwen2.5-72B-Instruct-GGUF --include "qwen2.5-72b-instruct-q2_k-*-of-00007.gguf" --local-dir Qwen2.5-72B-Instruct-GGUF
Q3_K_M33.0 GBfile size, 9 parts36.0 GB33.0 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf Qwen/Qwen2.5-72B-Instruct-GGUF:Q3_K_M
Download
hf download Qwen/Qwen2.5-72B-Instruct-GGUF --include "qwen2.5-72b-instruct-q3_k_m-*-of-00009.gguf" --local-dir Qwen2.5-72B-Instruct-GGUF
Q4_038.5 GBfile size, 11 parts41.5 GB38.5 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf Qwen/Qwen2.5-72B-Instruct-GGUF:Q4_0
Download
hf download Qwen/Qwen2.5-72B-Instruct-GGUF --include "qwen2.5-72b-instruct-q4_0-*-of-00011.gguf" --local-dir Qwen2.5-72B-Instruct-GGUF
Q4_K_M41.0 GBfile size, 12 parts44.0 GB41.0 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf Qwen/Qwen2.5-72B-Instruct-GGUF:Q4_K_M
Download
hf download Qwen/Qwen2.5-72B-Instruct-GGUF --include "qwen2.5-72b-instruct-q4_k_m-*-of-00012.gguf" --local-dir Qwen2.5-72B-Instruct-GGUF
Q5_046.9 GBfile size, 13 parts49.9 GB46.9 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf Qwen/Qwen2.5-72B-Instruct-GGUF:Q5_0
Download
hf download Qwen/Qwen2.5-72B-Instruct-GGUF --include "qwen2.5-72b-instruct-q5_0-*-of-00013.gguf" --local-dir Qwen2.5-72B-Instruct-GGUF
Q5_K_M48.1 GBfile size, 14 parts51.1 GB48.1 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf Qwen/Qwen2.5-72B-Instruct-GGUF:Q5_K_M
Download
hf download Qwen/Qwen2.5-72B-Instruct-GGUF --include "qwen2.5-72b-instruct-q5_k_m-*-of-00014.gguf" --local-dir Qwen2.5-72B-Instruct-GGUF
Q6_K55.7 GBfile size, 16 parts58.7 GB55.7 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf Qwen/Qwen2.5-72B-Instruct-GGUF:Q6_K
Download
hf download Qwen/Qwen2.5-72B-Instruct-GGUF --include "qwen2.5-72b-instruct-q6_k-*-of-00016.gguf" --local-dir Qwen2.5-72B-Instruct-GGUF
Q8_072.2 GBfile size, 21 parts75.2 GB72.2 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf Qwen/Qwen2.5-72B-Instruct-GGUF:Q8_0
Download
hf download Qwen/Qwen2.5-72B-Instruct-GGUF --include "qwen2.5-72b-instruct-q8_0-*-of-00021.gguf" --local-dir Qwen2.5-72B-Instruct-GGUF
F16135.9 GBfile size, 42 parts138.9 GB135.9 + 2.5 + 0.50
Commands
llama.cpp
llama-server -hf Qwen/Qwen2.5-72B-Instruct-GGUF:F16
Download
hf download Qwen/Qwen2.5-72B-Instruct-GGUF --include "qwen2.5-72b-instruct-fp16-*-of-00042.gguf" --local-dir Qwen2.5-72B-Instruct-GGUF

MLX for Apple silicon

FormatSizeMemory neededRepo
MLX 3bitmlx-community/Qwen2.5-72B-Instruct-3bit29.6 GBfile size32.6 GB29.6 + 2.5 + 0.50Download page
MLX 4bitmlx-community/Qwen2.5-72B-Instruct-4bit38.1 GBfile size41.1 GB38.1 + 2.5 + 0.50Download page

Original weights size from the safetensors header, for vLLM, SGLang, Transformers

FormatSizeMemory neededRepo
BF16original weights135.4 GBfile size138.4 GB135.4 + 2.5 + 0.50
hf download Qwen/Qwen2.5-72B-Instruct --local-dir Qwen2.5-72B-Instruct
Other conversions (22; sizes not read)

The Hub tags 97 quantised repos as conversions of this model; we read the most downloaded and show the maker's own first.

Run it by API

Hugging Face Inference Providers, fetched 19 Nov 2026, 13:58 UTC

2 providers serve this model through the Hugging Face router. Prices are US dollars per million tokens. Precision is what the provider reports; a lower precision than the original weights can change answers. Live first, then cheapest output.

ProviderInOutContextPrecisionTokens/sFirst tokenUptime 24hToolsJSON
DeepInfra$0.36$0.4032kfp8430.2 s98.2%yesyes
featherless-ai???not given???not givennot given

Through the router, with a Hugging Face token (the router bills your Hugging Face account). Change the name after the colon to pick another provider.

curl https://router.huggingface.co/v1/chat/completions \ -H "Authorization: Bearer $HF_TOKEN" -H "Content-Type: application/json" \ -d '{"model": "Qwen/Qwen2.5-72B-Instruct:deepinfra", "messages": [{"role": "user", "content": "Hello"}]}'

Not built yet: one key for every open model, with BlockRun, and our own tested-quality score per provider. Until then, precision and measured speed are the best guide to a provider here.

Fine-tunes of this model

from our snapshot of the Hub

Repos that declare this model as their base, most downloaded first. Listing is not a recommendation; unreviewed ones have not been checked.

Planned for this page Not built yet

Try it in private chat
A link to Solana AI's confidential chat for the models it serves. This model is not listed as served there.
Fine-tune it
Fine-tuning a few base models from this page and publishing the result here.
Creators and earnings
Pages for people who publish fine-tunes and adapters of this model, and what they earn.
Fund and bounties
Bounties for new open models, paid from a public fund. How the fund will work.
Contribute compute
How to give compute to open model work, with Akash.

Model data from the Hugging Face Hub 19 Nov 2026, 13:43 UTC. The repo. Found a mistake on this page? Tell us.