Qwen

Qwen2 72B Instruct

Qwen2 72B Instruct is an open-weights 72B-parameter model by Qwen. It is listed as in-progress on the local.ai leaderboard.

open weights
Parameters
72B
Benchmarks (local.ai snapshot)
τ²-bench
GAIA
GDPval
HuggingFace metadata · updated 2024-10-08
repo
Qwen/Qwen2-72B-Instruct
license
other
downloads
54,586
likes
718

Metadata refreshed from the HuggingFace API (snapshot 2026-08-15T15:53:42.405Z). Context length and long-form descriptions are the next pipeline stage.

Does it fit your hardware?

Detect or select your GPU first. Green = weights + context headroom, yellow = tight, grey = does not fit.

Q4_K_M 41.8GQ6_K 64.8GQ8_0 79.2GFP16 144G
Quantization≈ File sizeNote
IQ2_XXS19.4 GBExtreme compression, largest quality loss
IQ3_XXS25.9 GBVery small
Q3_K_M33.1 GBSmall
NVFP436 GBNVIDIA 4-bit
Q4_K_S38.9 GBCommon
Q4_K_M41.8 GBMost common sweet spot
Q5_K_M54.7 GBBalanced
Q6_K64.8 GBNear-lossless
Q8_079.2 GBNear-lossless
FP16144 GBOriginal weights

Actual files on HuggingFace

FileSize
model-00006-of-00037.safetensors3.72 GB
model-00010-of-00037.safetensors3.72 GB
model-00014-of-00037.safetensors3.72 GB
model-00018-of-00037.safetensors3.72 GB
model-00022-of-00037.safetensors3.72 GB
model-00026-of-00037.safetensors3.72 GB
GGUF: bartowski/Qwen2.5-72B-Instruct-GGUF · 65,924 downloads
FileSize
Qwen2.5-72B-Instruct-IQ1_M.gguf22.11 GBdownload
Qwen2.5-72B-Instruct-IQ2_M.gguf27.32 GBdownload
Qwen2.5-72B-Instruct-IQ2_XS.gguf25.20 GBdownload
Qwen2.5-72B-Instruct-IQ2_XXS.gguf23.74 GBdownload
Qwen2.5-72B-Instruct-IQ3_M.gguf33.07 GBdownload
Qwen2.5-72B-Instruct-IQ3_XXS.gguf29.66 GBdownload
Qwen2.5-72B-Instruct-IQ4_XS.gguf36.98 GBdownload
Qwen2.5-72B-Instruct-Q2_K.gguf27.76 GBdownload
Qwen2.5-72B-Instruct-Q2_K_L.gguf28.90 GBdownload
Qwen2.5-72B-Instruct-Q3_K_L.gguf36.79 GBdownload
Qwen2.5-72B-Instruct-Q3_K_M.gguf35.11 GBdownload
Qwen2.5-72B-Instruct-Q3_K_S.gguf32.12 GBdownload
GGUF: mradermacher/Qwen2.5-72B-Instruct-heretic-i1-GGUF · 9,327 downloads
FileSize
Qwen2.5-72B-Instruct-heretic.i1-IQ1_M.gguf22.11 GBdownload
Qwen2.5-72B-Instruct-heretic.i1-IQ1_S.gguf21.13 GBdownload
Qwen2.5-72B-Instruct-heretic.i1-IQ2_M.gguf27.32 GBdownload
Qwen2.5-72B-Instruct-heretic.i1-IQ2_S.gguf26.02 GBdownload
Qwen2.5-72B-Instruct-heretic.i1-IQ2_XS.gguf25.20 GBdownload
Qwen2.5-72B-Instruct-heretic.i1-IQ2_XXS.gguf23.74 GBdownload
Qwen2.5-72B-Instruct-heretic.i1-IQ3_M.gguf33.07 GBdownload
Qwen2.5-72B-Instruct-heretic.i1-IQ3_S.gguf32.12 GBdownload
Qwen2.5-72B-Instruct-heretic.i1-IQ3_XS.gguf30.59 GBdownload
Qwen2.5-72B-Instruct-heretic.i1-IQ3_XXS.gguf29.66 GBdownload
Qwen2.5-72B-Instruct-heretic.i1-IQ4_XS.gguf36.98 GBdownload
Qwen2.5-72B-Instruct-heretic.i1-Q2_K.gguf27.76 GBdownload

What you save self-hosting this model

Uses the model's own API price when it exists; otherwise the closest commercial equivalent.

Waiting for hardware detection…

RunLocal presents third-party leaderboard data (local.ai / Exo Labs, 2026-08) with its own fit and cost estimates. Verify pricing, licenses and file sizes at the source before deploying a model.