Actual Models

Models recommended by Actual

Curated GGUF releases prepared for private inference on your own computers.

13 models25 variants

Updated

View registry JSON

Convai Innovations1 variantConverted

Laya Multilingual — FP16 decision model

QuantizationF16
Download628.3MB
Estimated memory722.5MB
run model
Technical details
Model ID
convaiinnovations/laya-multilingual@e4e9ddf21a7b1903b7acffd8814ad4307bf63a67
Owner
Actual Computer
License
apache-2.0
Source revision
convaiinnovations/laya-multilingual@e4e9ddf21a7b1903b7acffd8814ad4307bf63a67

Variant checksums

  • F167a0b2b058dceaf21a20cf6eef3837b3eeb4c7c3535e3722181f051fcf55ff574

Google2 variantsBF16 weightsConversion date unavailable

Gemma 4 E2B IT

QuantizationQ8_0
Download5.5GB
Estimated memory6.4GB
run model
QuantizationQ4_K_M
Download4.1GB
Estimated memory4.7GB
run model
Native weights (SafeTensors)Links
FormatBF16
Download9.6GB
Sourcegemma-4-E2B-it@905e84b5
filesSHA256SUMS
Technical details
Model ID
google/gemma-4-e2b-it@905e84b50c4d2a365ebde34e685027578e6728db
Owner
Actual Computer
License
apache-2.0
Source revision
google/gemma-4-E2B-it@905e84b50c4d2a365ebde34e685027578e6728db

Variant checksums

  • Q8_0e527942d0d2e904f2588beddc7fc901bcd8c6171713f92d6cd0cbe1071f68237
  • Q4_K_Md576a86e7c21999bf78263253c87ab12c1375d04e6e6b2bb885cdfe971e0d5fb
  • BF16aa07fd3380ca3f3012de6d20e7d18df57a31fa646e9dff7f57ca165f5644419d

Google2 variantsBF16 + FP8 weightsConversion date unavailable

Gemma 4 E4B IT

QuantizationQ8_0
Download8.4GB
Estimated memory9.7GB
run model
QuantizationQ4_K_M
Download5.9GB
Estimated memory6.8GB
run model
Native weights (SafeTensors)Links
FormatBF16
Download14.9GB
Sourcegemma-4-E4B-it@d6436b3d
filesSHA256SUMS
FormatFP8
Download11.2GB
Sourcegemma-4-E4B-it@d6436b3d
filesSHA256SUMS
Technical details
Model ID
google/gemma-4-e4b-it@d6436b3d62967e1af08bbb046c6300b2a9ae8e85
Owner
Actual Computer
License
apache-2.0
Source revision
google/gemma-4-E4B-it@d6436b3d62967e1af08bbb046c6300b2a9ae8e85

Variant checksums

  • Q8_0ecacce4e487055edbc5feb435e1ca383a9d39a97aeb3e698125df626aeb820dd
  • Q4_K_M0d829a7f55dde54750d88cb3b5fed5b6af001605bbf220a9d1bdbf554a2bb34a
  • BF16011dc546a000c82e9649f529108b55d78f9253706d27c4e3e87ad00e307e5752
  • FP88d99613008ff65c7219f34d607ed0dabed334b8041a9a19b3809e767809dda32

Google2 variantsBF16 + FP8 weightsConversion date unavailable

Gemma 4 12B IT

QuantizationQ8_0
Download12.0GB
Estimated memory13.8GB
run model
QuantizationQ4_K_M
Download7.0GB
Estimated memory8.1GB
run model
Native weights (SafeTensors)Links
FormatBF16
Download22.3GB
Sourcegemma-4-12B-it@5926caa4
filesSHA256SUMS
FormatFP8
Download12.1GB
Sourcegemma-4-12B-it@5926caa4
filesSHA256SUMS
Technical details
Model ID
google/gemma-4-12b-it@5926caa4ec0cac5cbfadaf4077420520de1d5205
Owner
Actual Computer
License
apache-2.0
Source revision
google/gemma-4-12B-it@5926caa4ec0cac5cbfadaf4077420520de1d5205

Variant checksums

  • Q8_0c50e5032c4b8e347d5ba28f3a41e23f04b9f75aefc48d50d9b67962f1a049231
  • Q4_K_Mb9d2a464e3718a5a96f950c129745a9964a4d35ee263d6ef14916242bce3cfbb
  • BF161070b849b46e6d9337e2c277b0c2840af38491ff6ef75df918518530383dc62f
  • FP8a021ca4e834fbf5621b5bd4db47225d39bd572a2db520d8aef70505d06e5960f

Google2 variantsBF16 weightsConversion date unavailable

Gemma 4 26B A4B IT

QuantizationQ8_0
Download26.1GB
Estimated memory30.0GB
run model
QuantizationQ4_K_M
Download16.8GB
Estimated memory19.3GB
run model
Native weights (SafeTensors)Links
FormatBF16
Download48.1GB
Sourcegemma-4-26B-A4B-it@6e6f6ede
filesSHA256SUMS
Technical details
Model ID
google/gemma-4-26b-a4b-it@6e6f6edea8c52db2094dca3086e4b963a0034dfc
Owner
Actual Computer
License
apache-2.0
Source revision
google/gemma-4-26B-A4B-it@6e6f6edea8c52db2094dca3086e4b963a0034dfc

Variant checksums

  • Q8_0a373ef67674f0dd2a6ac1b7bf04fc84f22a4a426a60dab4a51c9c22534db79e9
  • Q4_K_Me7a9c0ba7de460d194941f5128fe4bc912b3ebb1b6725e799d89429266a6aa4a
  • BF16960731550334b226f50175a78f1d798c8ab62ded8f0bd2a4a51c90f4177afc46

Qwen2 variantsFP8 + BF16 weightsConversion date unavailable

Qwen3.8 27B

QuantizationQ8_0
Download30.9GB
Estimated memory35.5GB
run model
QuantizationQ4_K_M
Download19.5GB
Estimated memory22.4GB
run model
Native weights (SafeTensors)Links
FormatFP8
Download28.8GB
SourceQwen3.8-27B-FP8@017b9c7a
filesSHA256SUMS
FormatBF16
Download51.8GB
SourceQwen3.8-27B@1d4bf0f2
filesSHA256SUMS
Technical details
Model ID
qwen/qwen3.8-27b@1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0
Owner
Actual Computer
License
apache-2.0
Source revision
Qwen/Qwen3.8-27B@1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0

Variant checksums

  • Q8_0b1f5ffc6b165f40ede4c3656857f81dfbca78582086034e649472ee4010d80c3
  • Q4_K_Mc6054f4fe3502fcd96ee28150c6c6c0f4a1fb9889bd1b8b77d4d6193ab58f9a8
  • FP89db38d25d8caae8978bc6d3e33b6b7700b41a1b629685dbf00da58d8862b600a
  • BF16366f666e2fa4696a14497cfee8735ace13d1e8876cf600f944817545ef99ff50

Meta2 variantsBF16 + FP8 weightsConversion date unavailable

Muse Glimmer 30B

QuantizationQ8_0
Download31.2GB
Estimated memory35.8GB
run model
QuantizationQ4_K_M
Download19.4GB
Estimated memory22.3GB
run model
Native weights (SafeTensors)Links
FormatBF16
Download55.5GB
SourceMuse-Glimmer-30B@97c77dff
filesSHA256SUMS
FormatFP8
Download32.1GB
SourceMuse-Glimmer-30B@97c77dff
filesSHA256SUMS
Technical details
Model ID
meta-models/muse-glimmer-30b@97c77dff50b2797bcc558fa2d909761dbc575c59
Owner
Actual Computer
License
apache-2.0
Source revision
meta-models/Muse-Glimmer-30B@97c77dff50b2797bcc558fa2d909761dbc575c59

Variant checksums

  • Q8_02998873e80edd387960dd63c41c7d79bd2b992beff4b5caf03c5c0b144d4f3a4
  • Q4_K_M984316a5bd706d417f0feb0c23051497a0fc1450283fc559668610ed909743c9
  • BF16dbbef5908ae8f2470bcd8e3b702a4525b7d78f312bb5a6dd42eb05906cf134bb
  • FP8627be7d2e2ee7b65a09376c063fbafc425ed85619899e9781080daa882640323

Google2 variantsBF16 + FP8 weightsConversion date unavailable

Gemma 4 31B IT

QuantizationQ8_0
Download31.5GB
Estimated memory36.2GB
run model
QuantizationQ4_K_M
Download18.5GB
Estimated memory21.3GB
run model
Native weights (SafeTensors)Links
FormatBF16
Download58.3GB
Sourcegemma-4-31B-it@fb9ae262
filesSHA256SUMS
FormatFP8
Download31.0GB
Sourcegemma-4-31B-it@fb9ae262
filesSHA256SUMS
Technical details
Model ID
google/gemma-4-31b-it@fb9ae262347c3945692f09a612f8bb189def854f
Owner
Actual Computer
License
apache-2.0
Source revision
google/gemma-4-31B-it@fb9ae262347c3945692f09a612f8bb189def854f

Variant checksums

  • Q8_0f643e4ebd7d96f4c1b1a8292c5a69780fb317bf2327f2eceb2b2dfca38f46390
  • Q4_K_M42c4902d0c7ef86a70d820fcc6fbffaa9aa5fe29c2883fdcd82526f1ad6b3a96
  • BF167a9e8d6dc8a517286ea5363b88858ec61e67abfb2725e31d2a4d0ab4f4c0e83b
  • FP8d4e791ed45ec02672616d6d3e8d2a35fb5a8e97e8becab53184825dbbe2d7e1c

NVIDIA2 variantsBF16 + FP8 weightsConversion date unavailable

NVIDIA Nemotron 3.5 Lightning 30B A3B BF16

QuantizationQ8_0
Download32.6GB
Estimated memory37.5GB
run model
QuantizationQ4_K_M
Download23.7GB
Estimated memory27.2GB
run model
Native weights (SafeTensors)Links
FormatBF16
Download61.3GB
SourceNVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@33268dc8
filesSHA256SUMS
FormatFP8
Download32.6GB
SourceNVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@33268dc8
filesSHA256SUMS
Technical details
Model ID
nvidia/nvidia-nemotron-3.5-lightning-30b-a3b-bf16@33268dc8a6da85a56be2b12241453e4e1237bbe1
Owner
Actual Computer
License
OpenMDW-1.1
Source revision
nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16@33268dc8a6da85a56be2b12241453e4e1237bbe1

Variant checksums

  • Q8_07c28974abe94badc0e760467f7ab63bd52928a68cd70aad665635536e576bcd6
  • Q4_K_Mc015ca2eb6dbfeeb5e3a684d32a9b975572ae4c9c5e30b5beb112582cc00c584
  • BF16659a8ae1f7f6ae1a8c55696f38cc7af0f27fcf7b2b381a598ad0ae1b4ae5bac4
  • FP8e2a9874d744db2c702fc7d7343bafcdf76072a6eb4bb76297aa2376de5599d11

Nvidia2 variantsFP8 + BF16 weightsConversion date unavailable

Nemotron 3 Nano Omni 30B A3B Reasoning Bf16

QuantizationQ8_0
Download32.8GB
Estimated memory37.7GB
run model
QuantizationQ4_K_M
Download24.3GB
Estimated memory28.0GB
run model
Native weights (SafeTensors)Links
FormatFP8
Download32.8GB
SourceNemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8@0acd4b1c
filesSHA256SUMS
FormatBF16
Download61.5GB
SourceNemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16@87c1afb6
filesSHA256SUMS
Technical details
Model ID
nvidia/nemotron-3-nano-omni-30b-a3b-reasoning-bf16@87c1afb6dc3e5080345cc4e430146a44121541ac
Owner
Actual Computer
License
other
Source revision
nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16@87c1afb6dc3e5080345cc4e430146a44121541ac

Variant checksums

  • Q8_0e5f4d467d5b890916a07a3d92204a4031e461506a8eb0f5a74073415921cc52e
  • Q4_K_M2e8742b94af603a0e54e976b8342be6d865af15f8e2e4930495d7fff219c1996
  • FP8d53328cf6217d51162a95cd7aaefa1d4d9639e35f7294ebf458ff67cfc9e5cb6
  • BF16cc8e12331746e78283a5754d997074851135b1f0590583e38476a62b158f398e

nvidia2 variantsBF16 + NVFP4 weightsConverted

nvidia-nemotron-3.5-super-vl-120b-a12b-bf16

QuantizationQ8_0
Download124.2GB
Estimated memory142.8GB
run model
QuantizationQ4_K_M
Download84.0GB
Estimated memory96.6GB
run model
Native weights (SafeTensors)Links
FormatBF16
Download231.8GB
SourceNVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16@533d6623
filesSHA256SUMS
FormatNVFP4
Download69.9GB
SourceNVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4@7d4bc747
filesSHA256SUMS
Technical details
Model ID
nvidia/nvidia-nemotron-3.5-super-vl-120b-a12b-bf16@533d66230b6e75518bf32d2d7a7b95a24c3dd517
Owner
Actual Computer
License
other
Source revision
nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16@533d66230b6e75518bf32d2d7a7b95a24c3dd517

Variant checksums

  • Q8_0d73940ecf8fdb241a4c5d0a2d598027719b2c9a47ec37b6c4f49ce5855c41d8b
  • Q4_K_M99b832d35381feae805c5259939b97b396d0277601757a72bb3883ab7105d7d7
  • BF162b5a26066cf597785821da202e290f14250479b94655ab65ca3ee7a5f251f856
  • NVFP46bb1e89a3a70cc41d179d4fc8a89cd0ad81f35001d7477574d82a114a62fce02

DeepSeek2 variantsFP8 weightsConversion date unavailable

DeepSeek V4 Flash

QuantizationQ8_0
Download144.3GB
Estimated memory166.0GB
run model
QuantizationQ4_K_M
Download141.1GB
Estimated memory162.3GB
run model
Native weights (SafeTensors)Links
FormatFP8
Download155.4GB
SourceDeepSeek-V4-Flash-0731@7872f01b
filesSHA256SUMS
Technical details
Model ID
deepseek-ai/deepseek-v4-flash@7872f01b1d1fe23eabc4c98b48bffcef5a386062
Owner
Actual Computer
License
mit
Source revision
deepseek-ai/DeepSeek-V4-Flash-0731@7872f01b1d1fe23eabc4c98b48bffcef5a386062

Variant checksums

  • Q8_02c1cc0b4cc43b87cff4064561065b7b0dcb338edcc1c6b0517ca3f25bf4899a4
  • Q4_K_Mfb18f5d102731b4d5ffdf8607730f727ce1cf8747059437a2dce450a4bb57717
  • FP8a6e9f995737daa000c50c68e890ef5825abab4f53edde7d0a3a90bea12f98658

Qwen2 variantsConverted

qwen3.8-flash-next

QuantizationQ8_0
Download178.7GB
Estimated memory205.5GB
run model
QuantizationQ4_K_M
Download113.6GB
Estimated memory130.6GB
run model
Technical details
Model ID
qwen/qwen3.8-flash-next@f5d08274bafd880402bd16f5e3e6c514136ec06c
Owner
Actual Computer
License
other
Source revision
Qwen/Qwen3.8-Flash-Next@f5d08274bafd880402bd16f5e3e6c514136ec06c

Variant checksums

  • Q8_06cc086b79630cae5080df683a2be70201d1031465546167ab4a97f6b6d212091
  • Q4_K_M300de6672210d5c2d98f28529918d096687d22201013ad2d7c63e29720f21380