Google

Diffusiongemma 26B A4B It

Diffusiongemma 26B A4B It is an open-weights 26B-parameter model by Google. It holds an Intelligence score of 50.3 on the local.ai leaderboard (2026-08 snapshot).

MoEdiffusionopen weights
50.3
local.ai Intelligence
Parameters
26B
MoE · 4B active parameters
Benchmarks (local.ai snapshot)
τ²-bench
66%
GAIA
45%
GDPval
51%
HuggingFace metadata · updated 2026-07-15
repo
google/diffusiongemma-26B-A4B-it
license
apache-2.0
downloads
2,029,745
likes
1,175

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 15.1GQ6_K 23.4GQ8_0 28.6GFP16 52G
Quantization≈ File sizeNote
IQ2_XXS7 GBExtreme compression, largest quality loss
IQ3_XXS9.4 GBVery small
Q3_K_M12 GBSmall
NVFP413 GBNVIDIA 4-bit
Q4_K_S14 GBCommon
Q4_K_M15.1 GBMost common sweet spot
Q5_K_M19.8 GBBalanced
Q6_K23.4 GBNear-lossless
Q8_028.6 GBNear-lossless
FP1652 GBOriginal weights

Actual files on HuggingFace

FileSize
model-00005-of-00011.safetensors4.58 GB
model-00007-of-00011.safetensors4.58 GB
model-00009-of-00011.safetensors4.58 GB
model-00003-of-00011.safetensors4.58 GB
model-00006-of-00011.safetensors4.55 GB
model-00008-of-00011.safetensors4.55 GB
GGUF: unsloth/diffusiongemma-26B-A4B-it-GGUF · 33,661 downloads
FileSize
diffusiongemma-26B-A4B-it-BF16.gguf47.07 GBdownload
diffusiongemma-26B-A4B-it-Q4_K_M.gguf15.65 GBdownload
diffusiongemma-26B-A4B-it-Q5_K_M.gguf17.83 GBdownload
diffusiongemma-26B-A4B-it-Q6_K.gguf21.10 GBdownload
diffusiongemma-26B-A4B-it-Q8_0.gguf25.03 GBdownload
GGUF: yabaimimi/edwixx__diffusiongemma-26B-A4B-it-HERETIC-Uncensored-GGUF · 13,117 downloads
FileSize
edwixx__diffusiongemma-26B-A4B-it-HERETIC-Uncensored-BF16.gguf47.07 GBdownload
edwixx__diffusiongemma-26B-A4B-it-HERETIC-Uncensored-FP16.gguf47.07 GBdownload
edwixx__diffusiongemma-26B-A4B-it-HERETIC-Uncensored-IQ3_M.gguf11.55 GBdownload
edwixx__diffusiongemma-26B-A4B-it-HERETIC-Uncensored-IQ3_S.gguf11.39 GBdownload
edwixx__diffusiongemma-26B-A4B-it-HERETIC-Uncensored-IQ4_NL.gguf13.59 GBdownload
edwixx__diffusiongemma-26B-A4B-it-HERETIC-Uncensored-IQ4_XS.gguf13.11 GBdownload
edwixx__diffusiongemma-26B-A4B-it-HERETIC-Uncensored-Q2_K.gguf9.86 GBdownload
edwixx__diffusiongemma-26B-A4B-it-HERETIC-Uncensored-Q3_K_M.gguf12.38 GBdownload
edwixx__diffusiongemma-26B-A4B-it-HERETIC-Uncensored-Q4_K_M.gguf15.65 GBdownload
edwixx__diffusiongemma-26B-A4B-it-HERETIC-Uncensored-Q5_K_M.gguf17.83 GBdownload
edwixx__diffusiongemma-26B-A4B-it-HERETIC-Uncensored-Q6_K.gguf21.10 GBdownload
edwixx__diffusiongemma-26B-A4B-it-HERETIC-Uncensored-Q8_0.gguf25.03 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.