SAVRN
Search Contact SAVRN

Open-weight model

MiMo-V2.6-Pro-RL-ARVQ-hybrid

by Jarrel Seah jarrelscy/MiMo-V2.6-Pro-RL-ARVQ-hybrid

MiMo-V2.6-Pro-RL-ARVQ-hybrid is an open-weight model from Jarrel Seah, released under MIT License. It has 119B parameters and a 1,048,576-token context. At 16-bit it needs about 285.6 GB of GPU memory, which fits on 1x MI355X from $2.59 an hour; at 4-bit, 71.4 GB on 1x MI300X from $1.85, at the lowest prices in the SAVRN Index.

Complete uploaded 5% hot checkpoint: 1325 of 26496 routed experts use NVFP4. Remaining experts use per-expert ARVQ FP4 books and FP16 block scales.

Parameters119B
Context1,048,576
Weights322.7 GB
Licensemit
AccessOpen weights
Monthly Downloads—

Runs On

What it takes to serve MiMo-V2.6-Pro-RL-ARVQ-hybrid (119B 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 238.0 GB 285.6 GB 1x MI355X (288 GB)
Vultr
$2.59 2x MI300X $3.70 · 2x MI325X $4.00
8-bit 119.0 GB 142.8 GB 1x MI300X (192 GB)
Vultr
$1.85 1x MI325X $2.00 · 1x MI355X $2.59
4-bit 59.5 GB 71.4 GB 1x MI300X (192 GB)
Vultr
$1.85 1x H100 $1.99 · 1x MI325X $2.00

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.

MiMo-V2.6-Pro-RL-ARVQ-hybrid on every accelerator the SAVRN Index prices, at every precision

Model Card

By Jarrel Seah, published under mit, revision 67d6be01a93e.

Complete uploaded 5% hot checkpoint: 1325 of 26496 routed experts use NVFP4. Remaining experts use per-expert ARVQ FP4 books and FP16 block scales. Global allocation ranks routing-weighted output-error benefit measured on 65536 training-only calibration tokens. All 69 layers have prepared weights. Sequential PV is continuing through the remaining layers. Layers 1-20 reuse the earlier accepted all-cold candidates (layer 9 retained its initial fit). Their cold weight tensor payloads were verified identical to those underlying the cached inputs for layer 21. Layers 21 and 22 have completed PV; layers 23-69 follow sequentially, freezing the selected NVFP4 hot experts. Each accepted layer…

Read Jarrel Seah's full model card

MiMo-V2.6-Pro-RL ARVQ / NVFP4 hybrid

Complete uploaded 5% hot checkpoint: 1325 of 26496 routed experts use NVFP4. Remaining experts use per-expert ARVQ FP4 books and FP16 block scales. Global allocation ranks routing-weighted output-error benefit measured on 65536 training-only calibration tokens. All 69 layers have prepared weights.

Sequential PV is continuing through the remaining layers. Layers 1-20 reuse the earlier accepted all-cold candidates (layer 9 retained its initial fit). Their cold weight tensor payloads were verified identical to those underlying the cached inputs for layer 21. Layers 21 and 22 have completed PV; layers 23-69 follow sequentially, freezing the selected NVFP4 hot experts. Each accepted layer replaces its initial fit on HF as an atomic weights/roster/config commit. See pv_progress.json for verified upload status. Unfinished layers retain their already-uploaded initial fits. This is not yet a fully PV-completed or quality-qualified checkpoint.

All backbone, MTP, vision, audio, audio-tokenizer and DFlash weights are retained. Calibration is text-only with arithmetic emulation. Full-model quality evaluation is queued after remaining PV; SM120 serving and 1M-context memory fit remain unqualified. upload_verification.json identifies the snapshot whose hashes were verified; subsequent PV commits have their own per-layer hash checks.

Configuration

Architecture
MiMoV2ForCausalLM
Context length (tokens)
1,048,576
Layers
70
Hidden size
6,144
Feed-forward size
16,384
Attention heads
128
Key/value heads
8
Head dimension
192
Vocabulary size
152,576
Routed experts
384
Experts active per token
8
Sliding window (tokens)
128
RoPE base
10,000,000
Model type
mimo_v2
Quantization
nvfp4_arvq_hybrid

Identity and Version

Repository
jarrelscy/MiMo-V2.6-Pro-RL-ARVQ-hybrid
Publisher
Jarrel Seah
Task
Not stated by the source
Modality
Other
Library
vllm
Parameters
119B parameters
Languages
Not stated by the source
Revision
67d6be01a93e0c64e0ff1ade26e9978c4018ba45
First published
2026-09-22
Last updated
2026-09-23

Files and Weights

441 files, 322.7 GB in total. The weights are 213 files totalling 322.7 GB in pt, safetensors.

Weights213 files · 322.7 GB
Configuration217 files · 3.1 MB
Tokenizer5 files · 15.9 MB
Documentation1 file · 1.5 KB
Other4 files · 3.5 MB
Repository1 file · 1.7 KB
Every file
FileTypeSizeSHA-256
arvq-layer-001-down.safetensorsWeights1.3 GB a7cb7ebb8653
arvq-layer-001-gateup.safetensorsWeights2.6 GB 5693621b1834
arvq-layer-002-down.safetensorsWeights1.3 GB 6b6b17fdb1f0
arvq-layer-002-gateup.safetensorsWeights2.6 GB fd5262bf4643
arvq-layer-003-down.safetensorsWeights1.3 GB e8f710a52ec5
arvq-layer-003-gateup.safetensorsWeights2.6 GB 4b4f5112e3c9
arvq-layer-004-down.safetensorsWeights1.3 GB a2b7c69a5ef4
arvq-layer-004-gateup.safetensorsWeights2.6 GB 6ed97e6aaf39
arvq-layer-005-down.safetensorsWeights1.3 GB 6910dff93e09
arvq-layer-005-gateup.safetensorsWeights2.6 GB 5f0d74eb9e85
arvq-layer-006-down.safetensorsWeights1.3 GB 9625a50634b4
arvq-layer-006-gateup.safetensorsWeights2.6 GB beac26f68b08
arvq-layer-007-down.safetensorsWeights1.3 GB 288f964cc89f
arvq-layer-007-gateup.safetensorsWeights2.6 GB e8e0e5f88b63
arvq-layer-008-down.safetensorsWeights1.3 GB ee2d4b604821
arvq-layer-008-gateup.safetensorsWeights2.6 GB ac42841bfe7e
arvq-layer-009-down.safetensorsWeights1.3 GB 58622b1f6a84
arvq-layer-009-gateup.safetensorsWeights2.6 GB 9bd78e0093ff
arvq-layer-010-down.safetensorsWeights1.3 GB 8280cf3c6a50
arvq-layer-010-gateup.safetensorsWeights2.6 GB e0438ceb7e0a
arvq-layer-011-down.safetensorsWeights1.3 GB 98b01fd0d1af
arvq-layer-011-gateup.safetensorsWeights2.6 GB 82eb2cbe751a
arvq-layer-012-down.safetensorsWeights1.3 GB 379a87fdb18e
arvq-layer-012-gateup.safetensorsWeights2.6 GB 65f734c09ed3
arvq-layer-013-down.safetensorsWeights1.3 GB 34937e89a7a1
arvq-layer-013-gateup.safetensorsWeights2.6 GB 7d38708a866b
arvq-layer-014-down.safetensorsWeights1.3 GB 5a6bb7659d54
arvq-layer-014-gateup.safetensorsWeights2.6 GB 0e2ee947d69d
arvq-layer-015-down.safetensorsWeights1.3 GB ada24e6b7b36
arvq-layer-015-gateup.safetensorsWeights2.6 GB 491d8c2714e2
arvq-layer-016-down.safetensorsWeights1.3 GB 4bb81ca3f7e3
arvq-layer-016-gateup.safetensorsWeights2.6 GB 6ddefefd7635
arvq-layer-017-down.safetensorsWeights1.3 GB 4a1e4cade6eb
arvq-layer-017-gateup.safetensorsWeights2.6 GB 79a927bb7ac2
arvq-layer-018-down.safetensorsWeights1.3 GB a648ecd9ceb4
arvq-layer-018-gateup.safetensorsWeights2.6 GB 60828217f5f7
arvq-layer-019-down.safetensorsWeights1.3 GB 3607da2e0363
arvq-layer-019-gateup.safetensorsWeights2.6 GB 92538b002b7d
arvq-layer-020-down.safetensorsWeights1.3 GB 261fed9a49c9
arvq-layer-020-gateup.safetensorsWeights2.6 GB 1b4e0c748686
arvq-layer-021-down.safetensorsWeights1.3 GB a9802c2a7280
arvq-layer-021-gateup.safetensorsWeights2.6 GB e008fc57d35e
arvq-layer-022-down.safetensorsWeights1.3 GB b5784d931378
arvq-layer-022-gateup.safetensorsWeights2.6 GB 367946f65b6f
arvq-layer-023-down.safetensorsWeights1.3 GB 27d5819cdb71
arvq-layer-023-gateup.safetensorsWeights2.6 GB 81a077e678fb
arvq-layer-024-down.safetensorsWeights1.3 GB b05f39d5766a
arvq-layer-024-gateup.safetensorsWeights2.6 GB abce36759d4b
arvq-layer-025-down.safetensorsWeights1.3 GB d6fedf1d2f77
arvq-layer-025-gateup.safetensorsWeights2.6 GB 0ac1c69cef94
arvq-layer-026-down.safetensorsWeights1.3 GB d14044108e79
arvq-layer-026-gateup.safetensorsWeights2.6 GB aac4a9f77598
arvq-layer-027-down.safetensorsWeights1.3 GB bfec6ecab5ae
arvq-layer-027-gateup.safetensorsWeights2.6 GB bc6839c67a67
arvq-layer-028-down.safetensorsWeights1.3 GB 623b492b7ed3
arvq-layer-028-gateup.safetensorsWeights2.6 GB 6cbba5f6017a
arvq-layer-029-down.safetensorsWeights1.3 GB 8aa8d0f0a790
arvq-layer-029-gateup.safetensorsWeights2.6 GB 59d501b4b0c1
arvq-layer-030-down.safetensorsWeights1.3 GB badc83f71043
arvq-layer-030-gateup.safetensorsWeights2.6 GB 6e5f6bb4beef
arvq-layer-031-down.safetensorsWeights1.3 GB 6abe127c65b5
arvq-layer-031-gateup.safetensorsWeights2.6 GB 50c553b127e6
arvq-layer-032-down.safetensorsWeights1.3 GB a03465d1cdd3
arvq-layer-032-gateup.safetensorsWeights2.6 GB f936552b9e74
arvq-layer-033-down.safetensorsWeights1.3 GB afcd03a5d834
arvq-layer-033-gateup.safetensorsWeights2.6 GB 1eb7ce97b68c
arvq-layer-034-down.safetensorsWeights1.3 GB 6edc97246f0c
arvq-layer-034-gateup.safetensorsWeights2.6 GB 93b430994d77
arvq-layer-035-down.safetensorsWeights1.3 GB cdbe584dcd73
arvq-layer-035-gateup.safetensorsWeights2.6 GB 0e3685afacb0
arvq-layer-036-down.safetensorsWeights1.3 GB 6679689c2d27
arvq-layer-036-gateup.safetensorsWeights2.6 GB c8dcdb88964b
arvq-layer-037-down.safetensorsWeights1.3 GB b3ef4a1fa586
arvq-layer-037-gateup.safetensorsWeights2.6 GB 141867456199
arvq-layer-038-down.safetensorsWeights1.3 GB 904e5d7b5449
arvq-layer-038-gateup.safetensorsWeights2.6 GB 903cb018f624
arvq-layer-039-down.safetensorsWeights1.3 GB 30dc61b75dfb
arvq-layer-039-gateup.safetensorsWeights2.6 GB 038d0dcbe875
arvq-layer-040-down.safetensorsWeights1.3 GB 65dcbcf7aaf7
arvq-layer-040-gateup.safetensorsWeights2.6 GB 3f20afee61e6
arvq-layer-041-down.safetensorsWeights1.3 GB a62a13276dbe
arvq-layer-041-gateup.safetensorsWeights2.6 GB dc13b8130159
arvq-layer-042-down.safetensorsWeights1.3 GB be6f27c08ccf
arvq-layer-042-gateup.safetensorsWeights2.6 GB 7c139d353dc4
arvq-layer-043-down.safetensorsWeights1.3 GB 62cc4c7ad64e
arvq-layer-043-gateup.safetensorsWeights2.6 GB 4d8864a84291
arvq-layer-044-down.safetensorsWeights1.3 GB 3353821e4daa
arvq-layer-044-gateup.safetensorsWeights2.6 GB 511e35d00d4d
arvq-layer-045-down.safetensorsWeights1.3 GB 1d0f549a66ee
arvq-layer-045-gateup.safetensorsWeights2.6 GB 74b133c736b9
arvq-layer-046-down.safetensorsWeights1.3 GB 469347cdbbd8
arvq-layer-046-gateup.safetensorsWeights2.6 GB e3e3f567d858
arvq-layer-047-down.safetensorsWeights1.3 GB 4d142ba23c28
arvq-layer-047-gateup.safetensorsWeights2.5 GB 29bd44172240
arvq-layer-048-down.safetensorsWeights1.3 GB 824afe42243e
arvq-layer-048-gateup.safetensorsWeights2.5 GB 2655b3ac1064
arvq-layer-049-down.safetensorsWeights1.3 GB 6ef4585848b7
arvq-layer-049-gateup.safetensorsWeights2.5 GB 44a06ee90cd5
arvq-layer-050-down.safetensorsWeights1.3 GB 165c4da813d5
arvq-layer-050-gateup.safetensorsWeights2.5 GB 59a186cf0cba
arvq-layer-051-down.safetensorsWeights1.3 GB eaa4426c1bf6
arvq-layer-051-gateup.safetensorsWeights2.5 GB 5fb4428e0df9
arvq-layer-052-down.safetensorsWeights1.3 GB 1c4a12006e33
arvq-layer-052-gateup.safetensorsWeights2.5 GB c9c2ade8cc14
arvq-layer-053-down.safetensorsWeights1.2 GB 8d843353a1d7
arvq-layer-053-gateup.safetensorsWeights2.5 GB 0cebc9fd3591
arvq-layer-054-down.safetensorsWeights1.2 GB c7159033f92a
arvq-layer-054-gateup.safetensorsWeights2.5 GB f70541b0a7bc
arvq-layer-055-down.safetensorsWeights1.2 GB 97f39190e693
arvq-layer-055-gateup.safetensorsWeights2.4 GB 2c5db39e7dd8
arvq-layer-056-down.safetensorsWeights1.2 GB 76934f8e349b
arvq-layer-056-gateup.safetensorsWeights2.3 GB c0d2a721dcef
arvq-layer-057-down.safetensorsWeights1.2 GB 9b24d29afbbe
arvq-layer-057-gateup.safetensorsWeights2.3 GB 17afdca6ae63
arvq-layer-058-down.safetensorsWeights1.1 GB a7725e5d11ea
arvq-layer-058-gateup.safetensorsWeights2.2 GB 916be780cbc4
arvq-layer-059-down.safetensorsWeights1.1 GB 8cd8f2bf2418
arvq-layer-059-gateup.safetensorsWeights2.1 GB ecc89ba7187e
arvq-layer-060-down.safetensorsWeights1.0 GB 611295a9959f
arvq-layer-060-gateup.safetensorsWeights2.1 GB f950c33da5af
arvq-layer-061-down.safetensorsWeights1.1 GB f6b1ea71e260
arvq-layer-061-gateup.safetensorsWeights2.1 GB a279bae382a9
arvq-layer-062-down.safetensorsWeights1.0 GB a6cf061544c8
arvq-layer-062-gateup.safetensorsWeights2.0 GB 5fd7f7592dc0
arvq-layer-063-down.safetensorsWeights1.0 GB ca0686f797e5
arvq-layer-063-gateup.safetensorsWeights2.1 GB d1efdfd173aa
arvq-layer-064-down.safetensorsWeights973.2 MB 736ce2d86dc9
arvq-layer-064-gateup.safetensorsWeights1.9 GB 195e06fd6259
arvq-layer-065-down.safetensorsWeights936.4 MB 8b2841783815
arvq-layer-065-gateup.safetensorsWeights1.9 GB c5bbbf8cdf98
arvq-layer-066-down.safetensorsWeights993.3 MB e01a93e679b5
arvq-layer-066-gateup.safetensorsWeights2.0 GB 485ad3090858
arvq-layer-067-down.safetensorsWeights899.6 MB 8fd165de7ef0
arvq-layer-067-gateup.safetensorsWeights1.8 GB 7e688b8c704b
arvq-layer-068-down.safetensorsWeights812.7 MB 8ebc2500de32
arvq-layer-068-gateup.safetensorsWeights1.6 GB de75bcb1419d
arvq-layer-069-down.safetensorsWeights642.1 MB 9d28f23955b2
arvq-layer-069-gateup.safetensorsWeights1.3 GB b948f6e111c5
audio_tokenizer/model.safetensorsWeights1.9 GB 077033345d80
backbone-000.safetensorsWeights2.0 GB 4f5be09d843b
backbone-001.safetensorsWeights30.2 GB d857fa0d0b69
backbone-002.safetensorsWeights2.5 GB eba334c8613d
dflash/dflash_draft_model.safetensorsWeights5.5 GB 39208e3dd452
dflash/mask_embedding.ptWeights14.0 KB 436ad0885387
roster-layer-001.safetensorsWeights1.1 KB d278bbd42534
roster-layer-002.safetensorsWeights1.1 KB e908dbdf9b54
roster-layer-003.safetensorsWeights1.1 KB 66e901ba9948
roster-layer-004.safetensorsWeights1.1 KB 6b29a7d76127
roster-layer-005.safetensorsWeights1.1 KB 91cd8df7d23c
roster-layer-006.safetensorsWeights1.1 KB 60d263a85d77
roster-layer-007.safetensorsWeights1.1 KB f549fe09f38d
roster-layer-008.safetensorsWeights1.1 KB f7d0dd80f6f7
roster-layer-009.safetensorsWeights1.1 KB f10c4360ad5c
roster-layer-010.safetensorsWeights1.1 KB 7b913ed597a6
roster-layer-011.safetensorsWeights1.1 KB b0f2a6d497cc
roster-layer-012.safetensorsWeights1.1 KB 5fd25b75f327
roster-layer-013.safetensorsWeights1.1 KB 1bc52cdb8e8d
roster-layer-014.safetensorsWeights1.1 KB 253b9fcd69f8
roster-layer-015.safetensorsWeights1.1 KB aee4d0fdc93b
roster-layer-016.safetensorsWeights1.1 KB b314ae25ffce
roster-layer-017.safetensorsWeights1.1 KB 27661de6aeba
roster-layer-018.safetensorsWeights1.1 KB 94302e18316e
roster-layer-019.safetensorsWeights1.1 KB 7ea69924f7fa
roster-layer-020.safetensorsWeights1.1 KB db2aafff35ff
roster-layer-021.safetensorsWeights1.1 KB 1f4b3328bf9d
roster-layer-022.safetensorsWeights21.2 MB fb478fec04b1
roster-layer-023.safetensorsWeights1.1 KB d0b53f8cff31
roster-layer-024.safetensorsWeights1.1 KB f8420102e975
roster-layer-025.safetensorsWeights1.1 KB a2f4b05e9e40
roster-layer-026.safetensorsWeights1.1 KB b8ad43e49b54
roster-layer-027.safetensorsWeights1.1 KB 5d18992a0eba
roster-layer-028.safetensorsWeights1.1 KB 0116dac98d0b
roster-layer-029.safetensorsWeights1.1 KB 67341bf43879
roster-layer-030.safetensorsWeights1.1 KB e248982a484f
roster-layer-031.safetensorsWeights1.1 KB 74d6a61bbef6
roster-layer-032.safetensorsWeights1.1 KB e196c4f924b9
roster-layer-033.safetensorsWeights1.1 KB 2a329bc5a636
roster-layer-034.safetensorsWeights1.1 KB da6acf5af36a
roster-layer-035.safetensorsWeights1.1 KB eee7c3d17d91
roster-layer-036.safetensorsWeights1.1 KB df5ae603a1e0
roster-layer-037.safetensorsWeights1.1 KB 936690859cea
roster-layer-038.safetensorsWeights1.1 KB d61461ead6a9
roster-layer-039.safetensorsWeights1.1 KB be162e30f52c
roster-layer-040.safetensorsWeights1.1 KB 29c7383d05bd
roster-layer-041.safetensorsWeights1.1 KB 51bea8831409
roster-layer-042.safetensorsWeights1.1 KB 78629ed2c25c
roster-layer-043.safetensorsWeights42.5 MB 344355d45d52
roster-layer-044.safetensorsWeights21.2 MB 680fe80ba70a
roster-layer-045.safetensorsWeights42.5 MB 058351b0a222
roster-layer-046.safetensorsWeights1.1 KB 52252383263d
roster-layer-047.safetensorsWeights106.2 MB c514c94bec5b
roster-layer-048.safetensorsWeights148.6 MB 5ca02dd52bab
roster-layer-049.safetensorsWeights127.4 MB afebf4baa542
roster-layer-050.safetensorsWeights148.6 MB 60e8470781d8
roster-layer-051.safetensorsWeights212.3 MB b02bb66626b7
roster-layer-052.safetensorsWeights169.9 MB 98d3041a2b07
roster-layer-053.safetensorsWeights254.8 MB 3181e4651742
roster-layer-054.safetensorsWeights233.6 MB f644111f01a6
roster-layer-055.safetensorsWeights615.8 MB 13b4368048d3
roster-layer-056.safetensorsWeights806.9 MB 6da5cd5977f0
roster-layer-057.safetensorsWeights828.1 MB 626b04f934e8
roster-layer-058.safetensorsWeights1.2 GB e11a97b31282
roster-layer-059.safetensorsWeights1.4 GB 1ff314381495
roster-layer-060.safetensorsWeights1.6 GB 19c89145d1bb
roster-layer-061.safetensorsWeights1.4 GB bed1f64023df
roster-layer-062.safetensorsWeights1.7 GB 7f331c94d1ed
roster-layer-063.safetensorsWeights1.6 GB ba15b088f130
roster-layer-064.safetensorsWeights2.0 GB 38b4244cbd45
roster-layer-065.safetensorsWeights2.2 GB 008b9115d7b4
roster-layer-066.safetensorsWeights1.8 GB c4b6a791ee08
roster-layer-067.safetensorsWeights2.4 GB bbcb33a994cb
roster-layer-068.safetensorsWeights3.0 GB 837223040911
roster-layer-069.safetensorsWeights4.1 GB 2dfb310bbdf3
allocation.jsonConfiguration19.7 KB —
allocation_capture.jsonConfiguration816 B —
allocation_scores/layer1.jsonConfiguration14.8 KB —
allocation_scores/layer10.jsonConfiguration14.6 KB —
allocation_scores/layer11.jsonConfiguration14.6 KB —
allocation_scores/layer12.jsonConfiguration14.8 KB —
allocation_scores/layer13.jsonConfiguration14.6 KB —
allocation_scores/layer14.jsonConfiguration14.6 KB —
allocation_scores/layer15.jsonConfiguration14.5 KB —
allocation_scores/layer16.jsonConfiguration14.7 KB —
allocation_scores/layer17.jsonConfiguration14.1 KB —
allocation_scores/layer18.jsonConfiguration13.8 KB —
allocation_scores/layer19.jsonConfiguration13.0 KB —
allocation_scores/layer2.jsonConfiguration14.7 KB —
allocation_scores/layer20.jsonConfiguration12.9 KB —
allocation_scores/layer21.jsonConfiguration13.0 KB —
allocation_scores/layer22.jsonConfiguration13.0 KB —
allocation_scores/layer23.jsonConfiguration13.0 KB —
allocation_scores/layer24.jsonConfiguration13.0 KB —
allocation_scores/layer25.jsonConfiguration13.0 KB —
allocation_scores/layer26.jsonConfiguration13.0 KB —
allocation_scores/layer27.jsonConfiguration13.0 KB —
allocation_scores/layer28.jsonConfiguration13.0 KB —
allocation_scores/layer29.jsonConfiguration13.0 KB —
allocation_scores/layer3.jsonConfiguration14.7 KB —
allocation_scores/layer30.jsonConfiguration13.0 KB —
allocation_scores/layer31.jsonConfiguration12.9 KB —
allocation_scores/layer32.jsonConfiguration12.9 KB —
allocation_scores/layer33.jsonConfiguration13.0 KB —
allocation_scores/layer34.jsonConfiguration12.9 KB —
allocation_scores/layer35.jsonConfiguration12.9 KB —
allocation_scores/layer36.jsonConfiguration13.0 KB —
allocation_scores/layer37.jsonConfiguration12.9 KB —
allocation_scores/layer38.jsonConfiguration12.9 KB —
allocation_scores/layer39.jsonConfiguration12.9 KB —
allocation_scores/layer4.jsonConfiguration14.7 KB —
allocation_scores/layer40.jsonConfiguration12.9 KB —
allocation_scores/layer41.jsonConfiguration12.9 KB —
allocation_scores/layer42.jsonConfiguration12.9 KB —
allocation_scores/layer43.jsonConfiguration12.9 KB —
allocation_scores/layer44.jsonConfiguration12.9 KB —
allocation_scores/layer45.jsonConfiguration12.9 KB —
allocation_scores/layer46.jsonConfiguration12.9 KB —
allocation_scores/layer47.jsonConfiguration12.9 KB —
allocation_scores/layer48.jsonConfiguration12.9 KB —
allocation_scores/layer49.jsonConfiguration12.9 KB —
allocation_scores/layer5.jsonConfiguration14.6 KB —
allocation_scores/layer50.jsonConfiguration12.9 KB —
allocation_scores/layer51.jsonConfiguration12.9 KB —
allocation_scores/layer52.jsonConfiguration12.9 KB —
allocation_scores/layer53.jsonConfiguration12.9 KB —
allocation_scores/layer54.jsonConfiguration13.0 KB —
allocation_scores/layer55.jsonConfiguration12.9 KB —
allocation_scores/layer56.jsonConfiguration12.9 KB —
allocation_scores/layer57.jsonConfiguration12.9 KB —
allocation_scores/layer58.jsonConfiguration12.9 KB —
allocation_scores/layer59.jsonConfiguration12.9 KB —
allocation_scores/layer6.jsonConfiguration14.7 KB —
allocation_scores/layer60.jsonConfiguration13.0 KB —
allocation_scores/layer61.jsonConfiguration12.9 KB —
allocation_scores/layer62.jsonConfiguration12.9 KB —
allocation_scores/layer63.jsonConfiguration12.9 KB —
allocation_scores/layer64.jsonConfiguration12.9 KB —
allocation_scores/layer65.jsonConfiguration13.0 KB —
allocation_scores/layer66.jsonConfiguration12.9 KB —
allocation_scores/layer67.jsonConfiguration12.9 KB —
allocation_scores/layer68.jsonConfiguration12.9 KB —
allocation_scores/layer69.jsonConfiguration12.9 KB —
allocation_scores/layer7.jsonConfiguration14.6 KB —
allocation_scores/layer8.jsonConfiguration14.5 KB —
allocation_scores/layer9.jsonConfiguration14.6 KB —
audio_tokenizer/config.jsonConfiguration1.2 KB —
audio_tokenizer/generation_config.jsonConfiguration149 B —
campaign.jsonConfiguration1.2 KB —
cold_manifests/layer-001-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-002-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-003-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-004-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-005-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-006-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-007-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-008-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-009-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-010-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-011-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-012-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-013-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-014-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-015-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-016-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-017-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-018-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-019-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-020-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-021-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-022-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-023-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-024-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-025-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-026-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-027-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-028-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-029-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-030-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-031-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-032-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-033-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-034-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-035-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-036-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-037-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-038-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-039-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-040-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-041-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-042-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-043-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-044-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-045-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-046-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-047-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-048-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-049-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-050-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-051-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-052-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-053-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-054-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-055-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-056-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-057-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-058-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-059-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-060-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-061-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-062-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-063-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-064-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-065-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-066-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-067-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-068-manifest.jsonConfiguration4.3 KB —
cold_manifests/layer-069-manifest.jsonConfiguration4.3 KB —
config.jsonConfiguration131.5 KB —
configuration_mimo_v2.pyConfiguration10.0 KB —
dflash/config.jsonConfiguration1.2 KB —
dflash/dflash.pyConfiguration14.3 KB —
dflash/model.safetensors.index.jsonConfiguration4.7 KB —
evaluation/reference_sanity.jsonConfiguration191 B —
fit_completion.jsonConfiguration189 B —
fitting_config.jsonConfiguration1.6 KB —
generation_config.jsonConfiguration195 B —
hot_target.jsonConfiguration999 B —
model.safetensors.index.jsonConfiguration160.6 KB —
modeling_mimo_v2.pyConfiguration85.5 KB —
preparation_status.jsonConfiguration373 B —
preprocessor_config.jsonConfiguration350 B —
pv_progress.jsonConfiguration32.0 KB —
pv_reports/layer-001-selection.jsonConfiguration20.1 KB —
pv_reports/layer-001.jsonConfiguration18.7 KB —
pv_reports/layer-002-selection.jsonConfiguration30.7 KB —
pv_reports/layer-002.jsonConfiguration28.7 KB —
pv_reports/layer-003-selection.jsonConfiguration23.8 KB —
pv_reports/layer-003.jsonConfiguration22.2 KB —
pv_reports/layer-004-selection.jsonConfiguration24.9 KB —
pv_reports/layer-004.jsonConfiguration23.2 KB —
pv_reports/layer-005-selection.jsonConfiguration30.5 KB —
pv_reports/layer-005.jsonConfiguration28.4 KB —
pv_reports/layer-006-selection.jsonConfiguration21.4 KB —
pv_reports/layer-006.jsonConfiguration19.9 KB —
pv_reports/layer-007-selection.jsonConfiguration22.5 KB —
pv_reports/layer-007.jsonConfiguration20.9 KB —
pv_reports/layer-008-selection.jsonConfiguration30.7 KB —
pv_reports/layer-008.jsonConfiguration28.6 KB —
pv_reports/layer-009-selection.jsonConfiguration21.4 KB —
pv_reports/layer-009.jsonConfiguration19.9 KB —
pv_reports/layer-010-selection.jsonConfiguration30.7 KB —
pv_reports/layer-010.jsonConfiguration28.7 KB —
pv_reports/layer-011-selection.jsonConfiguration24.9 KB —
pv_reports/layer-011.jsonConfiguration23.2 KB —
pv_reports/layer-012-selection.jsonConfiguration21.3 KB —
pv_reports/layer-012.jsonConfiguration19.8 KB —
pv_reports/layer-013-selection.jsonConfiguration23.3 KB —
pv_reports/layer-013.jsonConfiguration21.7 KB —
pv_reports/layer-014-selection.jsonConfiguration30.4 KB —
pv_reports/layer-014.jsonConfiguration28.4 KB —
pv_reports/layer-015-selection.jsonConfiguration30.3 KB —
pv_reports/layer-015.jsonConfiguration28.3 KB —
pv_reports/layer-016-selection.jsonConfiguration21.3 KB —
pv_reports/layer-016.jsonConfiguration19.8 KB —
pv_reports/layer-017-selection.jsonConfiguration21.0 KB —
pv_reports/layer-017.jsonConfiguration19.5 KB —
pv_reports/layer-018-selection.jsonConfiguration20.9 KB —
pv_reports/layer-018.jsonConfiguration19.4 KB —
pv_reports/layer-019-selection.jsonConfiguration30.2 KB —
pv_reports/layer-019.jsonConfiguration28.1 KB —
pv_reports/layer-021-selection.jsonConfiguration30.2 KB —
pv_reports/layer-021.jsonConfiguration28.1 KB —
pv_reports/layer-022-selection.jsonConfiguration29.4 KB —
pv_reports/layer-022.jsonConfiguration27.4 KB —
pv_reports/layer-023-selection.jsonConfiguration30.1 KB —
pv_reports/layer-023.jsonConfiguration28.0 KB —
pv_reports/layer-024-selection.jsonConfiguration30.1 KB —
pv_reports/layer-024.jsonConfiguration28.0 KB —
pv_reports/layer-025-selection.jsonConfiguration30.1 KB —
pv_reports/layer-025.jsonConfiguration28.0 KB —
pv_reports/layer-026-selection.jsonConfiguration30.0 KB —
pv_reports/layer-026.jsonConfiguration28.0 KB —
pv_reports/layer-027-selection.jsonConfiguration30.0 KB —
pv_reports/layer-027.jsonConfiguration27.9 KB —
pv_reports/layer-028-selection.jsonConfiguration30.0 KB —
pv_reports/layer-028.jsonConfiguration27.9 KB —
pv_reports/layer-029-selection.jsonConfiguration27.9 KB —
pv_reports/layer-029.jsonConfiguration26.0 KB —
run_manifest.jsonConfiguration3.6 KB —
seed_ready.jsonConfiguration61 B —
upload_verification.jsonConfiguration254 B —
README.mdDocumentation1.5 KB —
MiMo_V2_6_technical_report.pdfOther3.0 MB fb81e6e08380
assets/architecture.pngOther405.4 KB d288768e1771
audio_tokenizer/chat_template.jinjaOther5.6 KB —
chat_template.jinjaOther3.9 KB —
.gitattributesRepository1.7 KB —
audio_tokenizer/tokenizer_config.jsonTokenizer6.1 KB —
merges.txtTokenizer1.7 MB —
tokenizer.jsonTokenizer11.4 MB ff15eb925890
tokenizer_config.jsonTokenizer12.2 KB —
vocab.jsonTokenizer2.8 MB —

License and Download

License
mit
Access
Open weights, no gate
Download size
322.7 GB
Download from Jarrel Seah

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

Built From

Memory Requirements

PrecisionWeights in memory
As published322.7 GB
16-bit238.0 GB
8-bit119.0 GB
4-bit59.5 GB

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

Built on This Model

Questions About MiMo-V2.6-Pro-RL-ARVQ-hybrid

How much GPU memory does MiMo-V2.6-Pro-RL-ARVQ-hybrid need?

About 285.6 GB at 16-bit and 71.4 GB at 4-bit: the weights (119B parameters) plus a working margin. A long context needs more.

What is the cheapest GPU to run MiMo-V2.6-Pro-RL-ARVQ-hybrid on?

At 16-bit, 1x MI355X from $2.59 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

Can I use MiMo-V2.6-Pro-RL-ARVQ-hybrid commercially?

Yes. MiMo-V2.6-Pro-RL-ARVQ-hybrid is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included.

What is MiMo-V2.6-Pro-RL-ARVQ-hybrid's context length?

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