SAVRN
Search Contact SAVRN

Open-weight model

GLM-5.3-Vision-NVFP4-ARVQ-v2-hybrid

by Jarrel Seah jarrelscy/GLM-5.3-Vision-NVFP4-ARVQ-v2-hybrid

GLM-5.3-Vision-NVFP4-ARVQ-v2-hybrid is an open-weight model from Jarrel Seah. It has 186.8B parameters and a 1,048,576-token context. At 16-bit it needs about 448.4 GB of GPU memory, which fits on 2x MI325X from $4.00 an hour; at 4-bit, 112.1 GB on 1x MI300X from $1.85, at the lowest prices in the SAVRN Index.

Per-expert ARVQ v3 cold experts, NVFP4 hot experts, ARVQ-scored REAP allocation. 75/75 MoE layers replaced by the full-corpus sequential PV campaign. Other layers retain their previously published weights; see pvprogress.json.

Parameters186.8B
Context1,048,576
Weights312.2 GB
License—
AccessOpen weights
Monthly Downloads—

Runs On

What it takes to serve GLM-5.3-Vision-NVFP4-ARVQ-v2-hybrid (186.8B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.

PrecisionWeightsMemory neededCheapest setupPer hourAlso fits
16-bit 373.7 GB 448.4 GB 2x MI325X (256 GB)
Vultr
$4.00 2x MI355X $5.18 · 3x MI300X $5.55
8-bit 186.8 GB 224.2 GB 1x MI325X (256 GB)
Vultr
$2.00 1x MI355X $2.59 · 2x MI300X $3.70
4-bit 93.4 GB 112.1 GB 1x MI300X (192 GB)
Vultr
$1.85 1x MI325X $2.00 · 1x MI355X $2.59

Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Oct 1, 2026.

GLM-5.3-Vision-NVFP4-ARVQ-v2-hybrid on every accelerator the SAVRN Index prices, at every precision

Model Card

Per-expert ARVQ v3 cold experts, NVFP4 hot experts, ARVQ-scored REAP allocation. 75/75 MoE layers replaced by the full-corpus sequential PV campaign. Other layers retain their previously published weights; see pvprogress.json. Each layer's two tensor files and reports are replaced together in one commit. Training draws sequentially from 18,001,846 text tokens at context 1024. Fixed validation and development-audit sets each contain 16,384 tokens. Adam trains FP4-constrained per-expert books and FP8-constrained per-block scales. The effective batch is 262,144 tokens, accumulated in four 65,536-token passes. Layers 4–26 use 69 updates: book/scale LR.048/.032 through update 45, then.012/.008.…

Excerpt from the card by Jarrel Seah.

Configuration

Architecture
Glm5vForConditionalGeneration
Context length (tokens)
1,048,576
Layers
78
Hidden size
6,144
Feed-forward size
12,288
Attention heads
64
Key/value heads
64
Head dimension
64
Vocabulary size
154,880
Routed experts
256
Experts active per token
8
Model type
glm5v
Quantization
nvfp4_arvq_hybrid

Identity and Version

Repository
jarrelscy/GLM-5.3-Vision-NVFP4-ARVQ-v2-hybrid
Publisher
Jarrel Seah
Task
Not stated by the source
Modality
Other
Library
vllm
Parameters
186.8B parameters
Languages
Not stated by the source
Revision
be8cd37751f667203448e5595df01ca0ef4f6000
First published
2026-09-20
Last updated
2026-09-21

Files and Weights

446 files, 312.2 GB in total. The weights are 274 files totalling 312.2 GB in safetensors.

Weights274 files · 312.2 GB
Configuration166 files · 3.4 MB
Tokenizer2 files · 20.2 MB
Documentation2 files · 19.5 KB
Other1 file · 10.7 KB
Repository1 file · 1.6 KB
Every file
FileTypeSizeSHA-256
arvq-layer-003-down.safetensorsWeights652.2 MB 5c58d04e9185
arvq-layer-003-gateup.safetensorsWeights1.3 GB b65c8df8d8c6
arvq-layer-004-down.safetensorsWeights581.9 MB ffff261c202f
arvq-layer-004-gateup.safetensorsWeights1.2 GB ca0187b62c25
arvq-layer-005-down.safetensorsWeights565.2 MB d2907ea1af7c
arvq-layer-005-gateup.safetensorsWeights1.1 GB fd4927ccb1e4
arvq-layer-006-down.safetensorsWeights531.8 MB acaf69635601
arvq-layer-006-gateup.safetensorsWeights1.1 GB c37d8686732c
arvq-layer-007-down.safetensorsWeights575.2 MB 30de6a3a9729
arvq-layer-007-gateup.safetensorsWeights1.2 GB 40a002ea072d
arvq-layer-008-down.safetensorsWeights705.7 MB e98424aae0a8
arvq-layer-008-gateup.safetensorsWeights1.4 GB 616642430941
arvq-layer-009-down.safetensorsWeights642.1 MB 237b5c5b4dec
arvq-layer-009-gateup.safetensorsWeights1.3 GB bde4abfce0df
arvq-layer-010-down.safetensorsWeights578.6 MB 69d315727e7c
arvq-layer-010-gateup.safetensorsWeights1.2 GB 58d0c079753f
arvq-layer-011-down.safetensorsWeights525.1 MB 21d84449b3be
arvq-layer-011-gateup.safetensorsWeights1.0 GB 89c4e17a34ea
arvq-layer-012-down.safetensorsWeights531.8 MB c0b53462169d
arvq-layer-012-gateup.safetensorsWeights1.1 GB 6df83b598560
arvq-layer-013-down.safetensorsWeights561.9 MB 62cbf844c187
arvq-layer-013-gateup.safetensorsWeights1.1 GB d7b87a5d8bb9
arvq-layer-014-down.safetensorsWeights535.1 MB 6298a2431141
arvq-layer-014-gateup.safetensorsWeights1.1 GB 0851075984b7
arvq-layer-015-down.safetensorsWeights605.3 MB edc418bae582
arvq-layer-015-gateup.safetensorsWeights1.2 GB b78414f2ef17
arvq-layer-016-down.safetensorsWeights555.2 MB 3b45c869739b
arvq-layer-016-gateup.safetensorsWeights1.1 GB 7d7414cd8b00
arvq-layer-017-down.safetensorsWeights588.6 MB 80c499b3a709
arvq-layer-017-gateup.safetensorsWeights1.2 GB 394904633ab8
arvq-layer-018-down.safetensorsWeights618.7 MB 2ffe22dbf766
arvq-layer-018-gateup.safetensorsWeights1.2 GB 88ca3d4384cf
arvq-layer-019-down.safetensorsWeights618.7 MB 21ff9efe5308
arvq-layer-019-gateup.safetensorsWeights1.2 GB 16d61211efe3
arvq-layer-020-down.safetensorsWeights645.5 MB d9e731354d8b
arvq-layer-020-gateup.safetensorsWeights1.3 GB d5ff7d1ca02c
arvq-layer-021-down.safetensorsWeights642.1 MB 2a131618b3f5
arvq-layer-021-gateup.safetensorsWeights1.3 GB 9420929ffd55
arvq-layer-022-down.safetensorsWeights645.5 MB 4b29152b4315
arvq-layer-022-gateup.safetensorsWeights1.3 GB 35ec3869d7e3
arvq-layer-023-down.safetensorsWeights568.5 MB 965624ab9296
arvq-layer-023-gateup.safetensorsWeights1.1 GB 38de065a38eb
arvq-layer-024-down.safetensorsWeights532.4 MB d705dfe05953
arvq-layer-024-gateup.safetensorsWeights1.1 GB 242d8c0822cb
arvq-layer-025-down.safetensorsWeights519.4 MB 241604c2e527
arvq-layer-025-gateup.safetensorsWeights1.0 GB 92b35e4c6ce0
arvq-layer-026-down.safetensorsWeights542.1 MB cba4ff1fc3fa
arvq-layer-026-gateup.safetensorsWeights1.1 GB cf3598c73ff8
arvq-layer-027-down.safetensorsWeights558.3 MB 51620286ed62
arvq-layer-027-gateup.safetensorsWeights1.1 GB f4d773431404
arvq-layer-028-down.safetensorsWeights551.8 MB 67163f4b2405
arvq-layer-028-gateup.safetensorsWeights1.1 GB c05942ff9517
arvq-layer-029-down.safetensorsWeights545.3 MB 1a4123d760bc
arvq-layer-029-gateup.safetensorsWeights1.1 GB 0b1c06620ed2
arvq-layer-030-down.safetensorsWeights551.8 MB 98adc04bfd7a
arvq-layer-030-gateup.safetensorsWeights1.1 GB d12a2796addf
arvq-layer-031-down.safetensorsWeights548.6 MB 476b482b1e52
arvq-layer-031-gateup.safetensorsWeights1.1 GB d58dfabaac99
arvq-layer-032-down.safetensorsWeights594.0 MB 74a9594a1cb2
arvq-layer-032-gateup.safetensorsWeights1.2 GB 1f69a13c0fa3
arvq-layer-033-down.safetensorsWeights581.0 MB 60a84fc1e146
arvq-layer-033-gateup.safetensorsWeights1.2 GB 71f0328629ec
arvq-layer-034-down.safetensorsWeights581.0 MB 45d9edfa624d
arvq-layer-034-gateup.safetensorsWeights1.2 GB eb7d5de9b107
arvq-layer-035-down.safetensorsWeights558.3 MB 781848c161cf
arvq-layer-035-gateup.safetensorsWeights1.1 GB 043c0dc60ae4
arvq-layer-036-down.safetensorsWeights555.1 MB 3c03d448675d
arvq-layer-036-gateup.safetensorsWeights1.1 GB 525739cc052a
arvq-layer-037-down.safetensorsWeights597.3 MB ea1238284979
arvq-layer-037-gateup.safetensorsWeights1.2 GB 038cc0f247df
arvq-layer-038-down.safetensorsWeights571.3 MB 6316ab1a977c
arvq-layer-038-gateup.safetensorsWeights1.1 GB 85d8ffc8af88
arvq-layer-039-down.safetensorsWeights574.6 MB e0d7d7f46304
arvq-layer-039-gateup.safetensorsWeights1.1 GB e46b43a84e17
arvq-layer-040-down.safetensorsWeights597.3 MB f34e4b4f37a1
arvq-layer-040-gateup.safetensorsWeights1.2 GB cb02a1a8e4c9
arvq-layer-041-down.safetensorsWeights600.5 MB dd4f5824c0d6
arvq-layer-041-gateup.safetensorsWeights1.2 GB 4248d095a3c6
arvq-layer-042-down.safetensorsWeights568.1 MB 89086aa2b90e
arvq-layer-042-gateup.safetensorsWeights1.1 GB 8b1fd1665aef
arvq-layer-043-down.safetensorsWeights600.5 MB d9524c47cc5b
arvq-layer-043-gateup.safetensorsWeights1.2 GB 386e0d0c865b
arvq-layer-044-down.safetensorsWeights561.6 MB 0bfa3ffe66a4
arvq-layer-044-gateup.safetensorsWeights1.1 GB bb7bcc5bf45f
arvq-layer-045-down.safetensorsWeights545.3 MB 41949dad7529
arvq-layer-045-gateup.safetensorsWeights1.1 GB e74f257dd138
arvq-layer-046-down.safetensorsWeights564.8 MB c266e9a8dbd1
arvq-layer-046-gateup.safetensorsWeights1.1 GB 46c428bdc897
arvq-layer-047-down.safetensorsWeights555.1 MB eb447e7c3d94
arvq-layer-047-gateup.safetensorsWeights1.1 GB ee9212cd50cd
arvq-layer-048-down.safetensorsWeights590.8 MB 379b19a0eef7
arvq-layer-048-gateup.safetensorsWeights1.2 GB 65856c12f673
arvq-layer-049-down.safetensorsWeights594.0 MB 79a998a4eb6a
arvq-layer-049-gateup.safetensorsWeights1.2 GB 061b62d959fd
arvq-layer-050-down.safetensorsWeights607.0 MB 718d50ff0bb9
arvq-layer-050-gateup.safetensorsWeights1.2 GB 27dabd5eadeb
arvq-layer-051-down.safetensorsWeights577.8 MB 1c0cd9c52ede
arvq-layer-051-gateup.safetensorsWeights1.2 GB 3ea53b1e16b9
arvq-layer-052-down.safetensorsWeights662.2 MB 3d307f8fa040
arvq-layer-052-gateup.safetensorsWeights1.3 GB 61af13b32a84
arvq-layer-053-down.safetensorsWeights590.8 MB 805cf490eb54
arvq-layer-053-gateup.safetensorsWeights1.2 GB 8c81a85680bd
arvq-layer-054-down.safetensorsWeights620.0 MB d2e1962776f1
arvq-layer-054-gateup.safetensorsWeights1.2 GB 38fae5f45a32
arvq-layer-055-down.safetensorsWeights620.0 MB 688f2b21ab05
arvq-layer-055-gateup.safetensorsWeights1.2 GB 7ddf0ededb31
arvq-layer-056-down.safetensorsWeights568.1 MB 2e98a6eb937a
arvq-layer-056-gateup.safetensorsWeights1.1 GB d59badfe4e33
arvq-layer-057-down.safetensorsWeights600.5 MB ed2d7b3d563c
arvq-layer-057-gateup.safetensorsWeights1.2 GB 5ffc581e410c
arvq-layer-058-down.safetensorsWeights610.3 MB 326beae92eed
arvq-layer-058-gateup.safetensorsWeights1.2 GB 5ca2e42de425
arvq-layer-059-down.safetensorsWeights646.0 MB 1e1b57931859
arvq-layer-059-gateup.safetensorsWeights1.3 GB aacb274e3fd1
arvq-layer-060-down.safetensorsWeights577.8 MB 4172f4a6177b
arvq-layer-060-gateup.safetensorsWeights1.2 GB 1daf13a86a01
arvq-layer-061-down.safetensorsWeights597.3 MB 82c3afb007c2
arvq-layer-061-gateup.safetensorsWeights1.2 GB 2c38a1288dc5
arvq-layer-062-down.safetensorsWeights616.8 MB b3f3c965a394
arvq-layer-062-gateup.safetensorsWeights1.2 GB 1501f844d26b
arvq-layer-063-down.safetensorsWeights581.0 MB c0e8359c9ec3
arvq-layer-063-gateup.safetensorsWeights1.2 GB 2d0be7a4c662
arvq-layer-064-down.safetensorsWeights590.8 MB 13893f160eb6
arvq-layer-064-gateup.safetensorsWeights1.2 GB 1b4aaadd386c
arvq-layer-065-down.safetensorsWeights629.7 MB 52afb5df3da9
arvq-layer-065-gateup.safetensorsWeights1.3 GB d6dc9cb65203
arvq-layer-066-down.safetensorsWeights646.0 MB 969ef1b6dca0
arvq-layer-066-gateup.safetensorsWeights1.3 GB ead3c2a88c74
arvq-layer-067-down.safetensorsWeights587.5 MB 1dcd8145b12d
arvq-layer-067-gateup.safetensorsWeights1.2 GB 2d242be41cdf
arvq-layer-068-down.safetensorsWeights623.2 MB 50d2a7f276ea
arvq-layer-068-gateup.safetensorsWeights1.2 GB 3178cfb374b8
arvq-layer-069-down.safetensorsWeights584.3 MB 45420a350920
arvq-layer-069-gateup.safetensorsWeights1.2 GB 2719fd6e18cd
arvq-layer-070-down.safetensorsWeights581.0 MB 52fb2965fea7
arvq-layer-070-gateup.safetensorsWeights1.2 GB 6e2fb29d646b
arvq-layer-071-down.safetensorsWeights590.8 MB 92cb0386f3fc
arvq-layer-071-gateup.safetensorsWeights1.2 GB c4b1fe66748c
arvq-layer-072-down.safetensorsWeights587.5 MB 0e43d6498364
arvq-layer-072-gateup.safetensorsWeights1.2 GB f7d3553edbcf
arvq-layer-073-down.safetensorsWeights584.3 MB 1612b3ac8cb1
arvq-layer-073-gateup.safetensorsWeights1.2 GB 25c14fc51c14
arvq-layer-074-down.safetensorsWeights548.6 MB f5ac130aa613
arvq-layer-074-gateup.safetensorsWeights1.1 GB 9fd825f3a4c5
arvq-layer-075-down.safetensorsWeights574.6 MB c29ffd505a6d
arvq-layer-075-gateup.safetensorsWeights1.1 GB 7708ebe4f64d
arvq-layer-076-down.safetensorsWeights607.0 MB 9241826f9711
arvq-layer-076-gateup.safetensorsWeights1.2 GB 46ad69dadb93
arvq-layer-077-down.safetensorsWeights600.5 MB d8847ad12cc8
arvq-layer-077-gateup.safetensorsWeights1.2 GB c69543825767
default-vision-00000.safetensorsWeights99.1 MB e7c6ce8c2742
default-vision-00082.safetensorsWeights833.8 MB af1c927919bd
hot-layer-003.safetensorsWeights1.3 GB 094b4ac96ede
hot-layer-004.safetensorsWeights1.6 GB 9b205c814659
hot-layer-005.safetensorsWeights2.1 GB 9d709b1a4290
hot-layer-006.safetensorsWeights2.4 GB c43d251a6b80
hot-layer-007.safetensorsWeights2.3 GB 28deacfd756c
hot-layer-008.safetensorsWeights169.9 MB 4e5ef47abf90
hot-layer-009.safetensorsWeights169.9 MB 4c01cdf10519
hot-layer-010.safetensorsWeights806.9 MB 95e9fd4dacc0
hot-layer-011.safetensorsWeights2.2 GB 42489fb0540e
hot-layer-012.safetensorsWeights1.9 GB 9c1e0c390027
hot-layer-013.safetensorsWeights2.3 GB 46b932531672
hot-layer-014.safetensorsWeights2.2 GB 36ecc9c95398
hot-layer-015.safetensorsWeights1.5 GB b803f34289d1
hot-layer-016.safetensorsWeights2.0 GB fdb873d5b1f2
hot-layer-017.safetensorsWeights1.2 GB 7f284b86a93d
hot-layer-018.safetensorsWeights1.4 GB 342e3a63fe75
hot-layer-019.safetensorsWeights1.1 GB 96472f59b256
hot-layer-020.safetensorsWeights1.1 GB 463752c7d933
hot-layer-021.safetensorsWeights1.1 GB f25cf2f22cf9
hot-layer-022.safetensorsWeights1.4 GB b1e4a8d55325
hot-layer-023.safetensorsWeights2.1 GB 7bc347f3469f
hot-layer-024.safetensorsWeights2.4 GB 7de7dcab19ad
hot-layer-025.safetensorsWeights2.2 GB bcc9918376f1
hot-layer-026.safetensorsWeights1.7 GB acc91b6b21b1
hot-layer-027.safetensorsWeights1.8 GB 68ba2a67023c
hot-layer-028.safetensorsWeights2.3 GB da24265743de
hot-layer-029.safetensorsWeights2.1 GB df71d9fd48f8
hot-layer-030.safetensorsWeights1.6 GB 135af5a70e89
hot-layer-031.safetensorsWeights1.7 GB 433b96bcc4a9
hot-layer-032.safetensorsWeights1.7 GB 8d9afddb122a
hot-layer-033.safetensorsWeights1.4 GB a8a1e62982d6
hot-layer-034.safetensorsWeights1.7 GB 0c8032e96fc4
hot-layer-035.safetensorsWeights1.9 GB 50c14593750d
hot-layer-036.safetensorsWeights2.0 GB f9d07fab50b3
hot-layer-037.safetensorsWeights1.5 GB b589ed73ce41
hot-layer-038.safetensorsWeights1.9 GB 718b2c929cae
hot-layer-039.safetensorsWeights1.8 GB ecd313b992aa
hot-layer-040.safetensorsWeights1.6 GB a466a9e55fa1
hot-layer-041.safetensorsWeights1.7 GB f9c0578ad0aa
hot-layer-042.safetensorsWeights1.8 GB e3d6eac70c3e
hot-layer-043.safetensorsWeights1.5 GB 4f24b2a4b0bc
hot-layer-044.safetensorsWeights1.9 GB 76755d97802c
hot-layer-045.safetensorsWeights1.8 GB 1be78206a65d
hot-layer-046.safetensorsWeights1.6 GB f87fa6c15645
hot-layer-047.safetensorsWeights1.7 GB befcdff5b5cb
hot-layer-048.safetensorsWeights1.5 GB fcc9010f0350
hot-layer-049.safetensorsWeights1.8 GB 5658c033f4e8
hot-layer-050.safetensorsWeights1.4 GB d7c67b974b91
hot-layer-051.safetensorsWeights1.5 GB 66d0eab06244
hot-layer-052.safetensorsWeights1.3 GB d32626ed8ce3
hot-layer-053.safetensorsWeights1.6 GB 26462bc4573e
hot-layer-054.safetensorsWeights1.8 GB 810cceea2364
hot-layer-055.safetensorsWeights1.4 GB e8905cd0e7f3
hot-layer-056.safetensorsWeights1.6 GB 769c2ba2a412
hot-layer-057.safetensorsWeights1.3 GB 467db3a28b9b
hot-layer-058.safetensorsWeights1.3 GB 94c9952a2301
hot-layer-059.safetensorsWeights1.6 GB 37ab1bdf9547
hot-layer-060.safetensorsWeights1.8 GB 20b878f115e6
hot-layer-061.safetensorsWeights1.6 GB bbd4bfd534d7
hot-layer-062.safetensorsWeights1.8 GB 68a67800d126
hot-layer-063.safetensorsWeights1.9 GB c981e384e747
hot-layer-064.safetensorsWeights1.7 GB bb7d5bc8614d
hot-layer-065.safetensorsWeights1.7 GB 0db5672bb51f
hot-layer-066.safetensorsWeights1.6 GB 8343a2770f92
hot-layer-067.safetensorsWeights2.0 GB 907aa89ef7ba
hot-layer-068.safetensorsWeights1.7 GB 25a05966accd
hot-layer-069.safetensorsWeights1.5 GB 3a0b4c07c6bf
hot-layer-070.safetensorsWeights1.5 GB 8db6425920a6
hot-layer-071.safetensorsWeights1.8 GB 2b6eaa9017a6
hot-layer-072.safetensorsWeights1.7 GB 79a7868a28e6
hot-layer-073.safetensorsWeights1.6 GB 19089307611f
hot-layer-074.safetensorsWeights1.8 GB 9b2a2d545236
hot-layer-075.safetensorsWeights1.1 GB ceef9c0ccfb4
hot-layer-076.safetensorsWeights1.2 GB 33b72896bb4f
hot-layer-077.safetensorsWeights212.3 MB 3ebb4f4b4af6
nvfp4-donor-00000.safetensorsWeights4.6 GB a43dc9d665a8
nvfp4-donor-00001.safetensorsWeights817.4 MB 5b4ea650f262
nvfp4-donor-00002.safetensorsWeights836.2 MB 6b29a5db388d
nvfp4-donor-00003.safetensorsWeights817.4 MB 36908da246f5
nvfp4-donor-00004.safetensorsWeights427.5 MB 258203cb24d2
nvfp4-donor-00005.safetensorsWeights817.4 MB 0c4725604992
nvfp4-donor-00006.safetensorsWeights836.2 MB abee295643c0
nvfp4-donor-00007.safetensorsWeights408.7 MB a8155f612b76
nvfp4-donor-00008.safetensorsWeights817.4 MB 60efad9153eb
nvfp4-donor-00009.safetensorsWeights836.2 MB e8b025327363
nvfp4-donor-00010.safetensorsWeights817.4 MB 93e6e7fe51ea
nvfp4-donor-00011.safetensorsWeights427.5 MB fb8fca71852d
nvfp4-donor-00012.safetensorsWeights817.4 MB 4188f9864586
nvfp4-donor-00013.safetensorsWeights836.2 MB 394568afc22f
nvfp4-donor-00014.safetensorsWeights408.7 MB ed9eb811f286
nvfp4-donor-00015.safetensorsWeights817.4 MB 4e736d020383
nvfp4-donor-00016.safetensorsWeights836.2 MB b76b20cafe14
nvfp4-donor-00017.safetensorsWeights616.1 MB 65a549a43e07
nvfp4-donor-00018.safetensorsWeights628.8 MB 5402c3a5720d
nvfp4-donor-00019.safetensorsWeights817.4 MB 9bbfc0703690
nvfp4-donor-00020.safetensorsWeights836.2 MB 89db367a548d
nvfp4-donor-00021.safetensorsWeights408.7 MB 5388995a03b0
nvfp4-donor-00022.safetensorsWeights817.4 MB f51a295d5492
nvfp4-donor-00023.safetensorsWeights836.2 MB d7e4bec830bb
nvfp4-donor-00024.safetensorsWeights408.7 MB cbbf63383502
nvfp4-donor-00025.safetensorsWeights836.2 MB c5521d8d29bc
nvfp4-donor-00026.safetensorsWeights817.4 MB be3e7b0661d2
nvfp4-donor-00027.safetensorsWeights836.2 MB b440e4e98193
nvfp4-donor-00028.safetensorsWeights408.7 MB 63cd0732ab32
nvfp4-donor-00029.safetensorsWeights817.4 MB eac4e320303e
nvfp4-donor-00030.safetensorsWeights836.2 MB 4e05aaacfa45
nvfp4-donor-00031.safetensorsWeights408.7 MB 051c2569acd1
nvfp4-donor-00032.safetensorsWeights836.2 MB 8ebda54179f8
nvfp4-donor-00033.safetensorsWeights817.4 MB e95ae24eed75
nvfp4-donor-00034.safetensorsWeights836.2 MB 9986d9b69895
nvfp4-donor-00035.safetensorsWeights408.7 MB 1d83c6bd3b53
nvfp4-donor-00036.safetensorsWeights817.4 MB 5e3c294434bc
nvfp4-donor-00037.safetensorsWeights836.2 MB 3d577b080e49
nvfp4-donor-00038.safetensorsWeights408.7 MB 30642d7f5dfa
nvfp4-donor-00039.safetensorsWeights836.2 MB 521de3bcf6a5
nvfp4-donor-00040.safetensorsWeights817.4 MB 42a8abf5e31b
nvfp4-donor-00041.safetensorsWeights616.1 MB 3d41fa2c3dfa
nvfp4-donor-00042.safetensorsWeights628.8 MB 40dcd2e49e3d
nvfp4-donor-00043.safetensorsWeights817.4 MB 89ef3e1924c2
nvfp4-donor-00044.safetensorsWeights6.8 GB 9a7d166efbf1
nvfp4-donor-00045.safetensorsWeights10.0 GB 0f6c729d555a
nvfp4-donor-00046.safetensorsWeights5.1 GB 8e45c3881367
arvq88_reference.pyConfiguration4.8 KB —
backbone_sources.jsonConfiguration139.4 KB —
btx_hybrid_manifest.jsonConfiguration706 B —
build_provenance.jsonConfiguration22.2 KB —
calibration_corpus.jsonConfiguration703 B —
cold_assignment.jsonConfiguration268.2 KB —
cold_manifests/layer-003-manifest.jsonConfiguration33.9 KB —
cold_manifests/layer-004-manifest.jsonConfiguration26.5 KB —
cold_manifests/layer-005-manifest.jsonConfiguration26.4 KB —
cold_manifests/layer-006-manifest.jsonConfiguration26.1 KB —
cold_manifests/layer-007-manifest.jsonConfiguration26.4 KB —
cold_manifests/layer-008-manifest.jsonConfiguration31.7 KB —
cold_manifests/layer-009-manifest.jsonConfiguration33.3 KB —
cold_manifests/layer-010-manifest.jsonConfiguration26.0 KB —
cold_manifests/layer-011-manifest.jsonConfiguration25.7 KB —
cold_manifests/layer-012-manifest.jsonConfiguration25.7 KB —
cold_manifests/layer-013-manifest.jsonConfiguration25.9 KB —
cold_manifests/layer-014-manifest.jsonConfiguration25.8 KB —
cold_manifests/layer-015-manifest.jsonConfiguration26.1 KB —
cold_manifests/layer-016-manifest.jsonConfiguration25.8 KB —
cold_manifests/layer-017-manifest.jsonConfiguration26.0 KB —
cold_manifests/layer-018-manifest.jsonConfiguration26.2 KB —
cold_manifests/layer-019-manifest.jsonConfiguration26.2 KB —
cold_manifests/layer-020-manifest.jsonConfiguration26.3 KB —
cold_manifests/layer-021-manifest.jsonConfiguration26.3 KB —
cold_manifests/layer-022-manifest.jsonConfiguration26.3 KB —
cold_manifests/layer-023-manifest.jsonConfiguration25.8 KB —
cold_manifests/layer-024-manifest.jsonConfiguration5.2 KB —
cold_manifests/layer-025-manifest.jsonConfiguration5.1 KB —
cold_manifests/layer-026-manifest.jsonConfiguration5.2 KB —
cold_manifests/layer-027-manifest.jsonConfiguration5.3 KB —
cold_manifests/layer-028-manifest.jsonConfiguration5.3 KB —
cold_manifests/layer-029-manifest.jsonConfiguration5.3 KB —
cold_manifests/layer-030-manifest.jsonConfiguration5.3 KB —
cold_manifests/layer-031-manifest.jsonConfiguration5.3 KB —
cold_manifests/layer-032-manifest.jsonConfiguration5.6 KB —
cold_manifests/layer-033-manifest.jsonConfiguration5.5 KB —
cold_manifests/layer-034-manifest.jsonConfiguration5.5 KB —
cold_manifests/layer-035-manifest.jsonConfiguration5.3 KB —
cold_manifests/layer-036-manifest.jsonConfiguration5.3 KB —
cold_manifests/layer-037-manifest.jsonConfiguration5.6 KB —
cold_manifests/layer-038-manifest.jsonConfiguration5.4 KB —
cold_manifests/layer-039-manifest.jsonConfiguration5.4 KB —
cold_manifests/layer-040-manifest.jsonConfiguration5.6 KB —
cold_manifests/layer-041-manifest.jsonConfiguration5.6 KB —
cold_manifests/layer-042-manifest.jsonConfiguration5.4 KB —
cold_manifests/layer-043-manifest.jsonConfiguration5.6 KB —
cold_manifests/layer-044-manifest.jsonConfiguration5.3 KB —
cold_manifests/layer-045-manifest.jsonConfiguration5.3 KB —
cold_manifests/layer-046-manifest.jsonConfiguration5.4 KB —
cold_manifests/layer-047-manifest.jsonConfiguration5.3 KB —
cold_manifests/layer-048-manifest.jsonConfiguration5.5 KB —
cold_manifests/layer-049-manifest.jsonConfiguration5.5 KB —
cold_manifests/layer-050-manifest.jsonConfiguration5.6 KB —
cold_manifests/layer-051-manifest.jsonConfiguration5.4 KB —
cold_manifests/layer-052-manifest.jsonConfiguration6.0 KB —
cold_manifests/layer-053-manifest.jsonConfiguration5.5 KB —
cold_manifests/layer-054-manifest.jsonConfiguration5.7 KB —
cold_manifests/layer-055-manifest.jsonConfiguration5.7 KB —
cold_manifests/layer-056-manifest.jsonConfiguration5.4 KB —
cold_manifests/layer-057-manifest.jsonConfiguration5.6 KB —
cold_manifests/layer-058-manifest.jsonConfiguration5.6 KB —
cold_manifests/layer-059-manifest.jsonConfiguration5.9 KB —
cold_manifests/layer-060-manifest.jsonConfiguration5.5 KB —
cold_manifests/layer-061-manifest.jsonConfiguration5.6 KB —
cold_manifests/layer-062-manifest.jsonConfiguration5.7 KB —
cold_manifests/layer-063-manifest.jsonConfiguration5.5 KB —
cold_manifests/layer-064-manifest.jsonConfiguration5.5 KB —
cold_manifests/layer-065-manifest.jsonConfiguration5.8 KB —
cold_manifests/layer-066-manifest.jsonConfiguration5.9 KB —
cold_manifests/layer-067-manifest.jsonConfiguration5.5 KB —
cold_manifests/layer-068-manifest.jsonConfiguration5.7 KB —
cold_manifests/layer-069-manifest.jsonConfiguration5.5 KB —
cold_manifests/layer-070-manifest.jsonConfiguration5.5 KB —
cold_manifests/layer-071-manifest.jsonConfiguration5.5 KB —
cold_manifests/layer-072-manifest.jsonConfiguration5.5 KB —
cold_manifests/layer-073-manifest.jsonConfiguration5.5 KB —
cold_manifests/layer-074-manifest.jsonConfiguration5.3 KB —
cold_manifests/layer-075-manifest.jsonConfiguration5.4 KB —
cold_manifests/layer-076-manifest.jsonConfiguration5.6 KB —
cold_manifests/layer-077-manifest.jsonConfiguration5.6 KB —
config.jsonConfiguration43.6 KB —
gates_report.jsonConfiguration17.4 KB —
generation_config.jsonConfiguration215 B —
kimi_k25_processor.pyConfiguration9.5 KB —
kimi_k25_vision_processing.pyConfiguration10.0 KB —
media_utils.pyConfiguration13.2 KB —
model.safetensors.index.jsonConfiguration293.9 KB —
preprocessor_config.jsonConfiguration685 B —
pv_layers/layer_003.jsonConfiguration26.2 KB —
pv_layers/layer_004.jsonConfiguration19.7 KB —
pv_layers/layer_005.jsonConfiguration19.6 KB —
pv_layers/layer_006.jsonConfiguration19.6 KB —
pv_layers/layer_007.jsonConfiguration19.5 KB —
pv_layers/layer_008.jsonConfiguration23.8 KB —
pv_layers/layer_009.jsonConfiguration25.6 KB —
pv_layers/layer_010.jsonConfiguration19.2 KB —
pv_layers/layer_011.jsonConfiguration19.1 KB —
pv_layers/layer_012.jsonConfiguration19.2 KB —
pv_layers/layer_013.jsonConfiguration19.1 KB —
pv_layers/layer_014.jsonConfiguration19.2 KB —
pv_layers/layer_015.jsonConfiguration19.1 KB —
pv_layers/layer_016.jsonConfiguration19.1 KB —
pv_layers/layer_017.jsonConfiguration19.1 KB —
pv_layers/layer_018.jsonConfiguration19.1 KB —
pv_layers/layer_019.jsonConfiguration19.1 KB —
pv_layers/layer_020.jsonConfiguration19.1 KB —
pv_layers/layer_021.jsonConfiguration19.1 KB —
pv_layers/layer_022.jsonConfiguration19.1 KB —
pv_layers/layer_023.jsonConfiguration19.0 KB —
pv_layers/layer_024.jsonConfiguration19.4 KB —
pv_layers/layer_025.jsonConfiguration19.5 KB —
pv_layers/layer_026.jsonConfiguration20.1 KB —
pv_layers/layer_027.jsonConfiguration18.0 KB —
pv_layers/layer_028.jsonConfiguration18.0 KB —
pv_layers/layer_029.jsonConfiguration22.8 KB —
pv_layers/layer_030.jsonConfiguration14.6 KB —
pv_layers/layer_031.jsonConfiguration23.6 KB —
pv_layers/layer_032.jsonConfiguration22.7 KB —
pv_layers/layer_033.jsonConfiguration24.4 KB —
pv_layers/layer_034.jsonConfiguration20.7 KB —
pv_layers/layer_035.jsonConfiguration23.5 KB —
pv_layers/layer_036.jsonConfiguration24.3 KB —
pv_layers/layer_037.jsonConfiguration22.6 KB —
pv_layers/layer_038.jsonConfiguration22.6 KB —
pv_layers/layer_039.jsonConfiguration22.6 KB —
pv_layers/layer_040.jsonConfiguration22.6 KB —
pv_layers/layer_041.jsonConfiguration22.6 KB —
pv_layers/layer_042.jsonConfiguration16.3 KB —
pv_layers/layer_043.jsonConfiguration24.2 KB —
pv_layers/layer_044.jsonConfiguration24.2 KB —
pv_layers/layer_045.jsonConfiguration24.1 KB —
pv_layers/layer_046.jsonConfiguration24.1 KB —
pv_layers/layer_047.jsonConfiguration24.2 KB —
pv_layers/layer_048.jsonConfiguration24.1 KB —
pv_layers/layer_049.jsonConfiguration22.5 KB —
pv_layers/layer_050.jsonConfiguration22.4 KB —
pv_layers/layer_051.jsonConfiguration22.5 KB —
pv_layers/layer_052.jsonConfiguration22.4 KB —
pv_layers/layer_053.jsonConfiguration24.1 KB —
pv_layers/layer_054.jsonConfiguration24.0 KB —
pv_layers/layer_055.jsonConfiguration24.0 KB —
pv_layers/layer_056.jsonConfiguration24.0 KB —
pv_layers/layer_057.jsonConfiguration24.0 KB —
pv_layers/layer_058.jsonConfiguration24.0 KB —
pv_layers/layer_059.jsonConfiguration24.0 KB —
pv_layers/layer_060.jsonConfiguration24.0 KB —
pv_layers/layer_061.jsonConfiguration24.0 KB —
pv_layers/layer_062.jsonConfiguration24.0 KB —
pv_layers/layer_063.jsonConfiguration24.0 KB —
pv_layers/layer_064.jsonConfiguration24.0 KB —
pv_layers/layer_065.jsonConfiguration24.1 KB —
pv_layers/layer_066.jsonConfiguration24.1 KB —
pv_layers/layer_067.jsonConfiguration24.1 KB —
pv_layers/layer_068.jsonConfiguration24.1 KB —
pv_layers/layer_069.jsonConfiguration24.1 KB —
pv_layers/layer_070.jsonConfiguration24.1 KB —
pv_layers/layer_071.jsonConfiguration24.1 KB —
pv_layers/layer_072.jsonConfiguration24.1 KB —
pv_layers/layer_073.jsonConfiguration24.1 KB —
pv_layers/layer_074.jsonConfiguration24.1 KB —
pv_layers/layer_075.jsonConfiguration24.1 KB —
pv_layers/layer_076.jsonConfiguration24.1 KB —
pv_layers/layer_077.jsonConfiguration23.2 KB —
pv_progress.jsonConfiguration2.5 KB —
source.jsonConfiguration270 B —
METHODOLOGY.mdDocumentation17.7 KB —
README.mdDocumentation1.8 KB —
chat_template.jinjaOther10.7 KB —
.gitattributesRepository1.6 KB —
tokenizer.jsonTokenizer20.2 MB 19e773648cb4
tokenizer_config.jsonTokenizer790 B —

License and Download

License
Not stated by the source
Access
Open weights, no gate
Download size
312.2 GB
Download from Jarrel Seah

Released by Jarrel Seah through its official repository on Hugging Face.

Memory Requirements

PrecisionWeights in memory
As published312.2 GB
16-bit373.7 GB
8-bit186.8 GB
4-bit93.4 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Questions About GLM-5.3-Vision-NVFP4-ARVQ-v2-hybrid

How much GPU memory does GLM-5.3-Vision-NVFP4-ARVQ-v2-hybrid need?

About 448.4 GB at 16-bit and 112.1 GB at 4-bit: the weights (186.8B parameters) plus a working margin. A long context needs more.

What is the cheapest GPU to run GLM-5.3-Vision-NVFP4-ARVQ-v2-hybrid on?

At 16-bit, 2x MI325X from $4.00 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

What is GLM-5.3-Vision-NVFP4-ARVQ-v2-hybrid's context length?

1,048,576 tokens, from the maximum position embeddings in its published configuration.