Skip to page content
SAVRN
Search Plan a project

Open-weight model

Qwen3-TTS-12Hz-0.6B-Zhongli-LoRa-32

by Rijul Paul rijulpaul/Qwen3-TTS-12Hz-0.6B-Zhongli-LoRa-32

Qwen3-TTS-12Hz-0.6B-Zhongli-LoRa-32 is an open-weight model from Rijul Paul. It has 906M parameters. At 16-bit it needs about 2.2 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index.

Parameters906M
Context—
Weights6.5 GB
License—
AccessOpen weights
Monthly Downloads—

Runs On

What it takes to serve Qwen3-TTS-12Hz-0.6B-Zhongli-LoRa-32 (906M 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 1.8 GB 2.2 GB 1x MI300X (192 GB)
Vultr		 $1.85 1x H100 $1.99 · 1x MI325X $2.00
8-bit 0.9 GB 1.1 GB 1x MI300X (192 GB)
Vultr		 $1.85 1x H100 $1.99 · 1x MI325X $2.00
4-bit 0.5 GB 0.5 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 23, 2026.

Qwen3-TTS-12Hz-0.6B-Zhongli-LoRa-32 on every accelerator the SAVRN Index prices, at every precision

Model Card

The publisher has not written a card for this model.

Configuration

Architecture
Qwen3TTSForConditionalGeneration
Model type
qwen3_tts

Identity and Version

Repository
rijulpaul/Qwen3-TTS-12Hz-0.6B-Zhongli-LoRa-32
Publisher
Rijul Paul
Task
Not stated by the source
Modality
Other
Library
Not stated by the source
Parameters
906M parameters
Languages
Not stated by the source
Revision
56e02f46c276b96c61b4a5d981211118d3fb179d
First published
2026-09-27
Last updated
2026-09-27

Files and Weights

97 files, 6.5 GB in total. The weights are 6 files totalling 6.5 GB in pt, safetensors.

Weights6 files · 6.5 GB
Configuration9 files · 42.1 KB
Tokenizer3 files · 4.5 MB
Other78 files · 21.4 MB
Repository1 file · 5.7 KB
Every file
FileTypeSizeSHA-256
adapter/adapter_model.safetensorsWeights47.6 MB cbba12d4a2d7
model.safetensorsWeights3.6 GB b9d076b50b37
speech_tokenizer/model.safetensorsWeights682.3 MB 836b7b357f5e
training-checkpoint/pytorch_model/bf16_zero_pp_rank_0_mp_rank_00_optim_states.ptWeights142.8 MB 84bb484a8779
training-checkpoint/pytorch_model/bf16_zero_pp_rank_1_mp_rank_00_optim_states.ptWeights142.8 MB 16bb7a12647e
training-checkpoint/pytorch_model/mp_rank_00_model_states.ptWeights1.9 GB 2f1c83b39b71
adapter/adapter_config.jsonConfiguration1.3 KB —
config.jsonConfiguration4.5 KB —
generation_config.jsonConfiguration245 B —
preprocessor_config.jsonConfiguration127 B —
speech_tokenizer/config.jsonConfiguration2.3 KB —
speech_tokenizer/configuration.jsonConfiguration76 B —
speech_tokenizer/preprocessor_config.jsonConfiguration234 B —
training-checkpoint/train_info.jsonConfiguration64 B —
training-checkpoint/zero_to_fp32.pyConfiguration33.3 KB —
samples/000/000.wavOther272.7 KB bef2009af09a
samples/000/001.wavOther318.8 KB db2c805d3110
samples/000/002.wavOther295.7 KB 2c2ca4594022
samples/000/003.wavOther410.9 KB 5c0d9f1946cc
samples/000/004.wavOther130.6 KB 78e9f9681959
samples/001/000.wavOther330.3 KB 2e699d00b233
samples/001/001.wavOther307.2 KB 3a5c7bf51eae
samples/001/002.wavOther257.3 KB 6a467c8c9d6f
samples/001/003.wavOther387.9 KB 6a27b8efe977
samples/001/004.wavOther134.4 KB e4b520deea9c
samples/002/000.wavOther311.1 KB 142ffecf1b74
samples/002/001.wavOther338.0 KB 2a9ecb674ac8
samples/002/002.wavOther276.5 KB 524afdc8efe4
samples/002/003.wavOther364.8 KB dbce26670ae0
samples/002/004.wavOther134.4 KB 42529971a6b2
samples/003/000.wavOther291.9 KB 435e1c038158
samples/003/001.wavOther272.7 KB b7fa849269fa
samples/003/002.wavOther242.0 KB c9a8e5eae374
samples/003/003.wavOther318.8 KB 0bb31c7f3977
samples/003/004.wavOther134.4 KB cfe6fe88bdb3
samples/004/000.wavOther311.1 KB 84325868f190
samples/004/001.wavOther353.3 KB e33dc7fb2f16
samples/004/002.wavOther303.4 KB 7f66666b620c
samples/004/003.wavOther380.2 KB 0f4e656ae299
samples/004/004.wavOther138.3 KB 818e4e650ca9
samples/005/000.wavOther334.1 KB f9d985b29d70
samples/005/001.wavOther299.6 KB e26ce342fd5f
samples/005/002.wavOther253.5 KB 7e3a53d795b9
samples/005/003.wavOther372.5 KB 462a611dadd0
samples/005/004.wavOther126.8 KB be2f79563ef9
samples/006/000.wavOther353.3 KB f99de39ebabd
samples/006/001.wavOther322.6 KB 41e54db06cd6
samples/006/002.wavOther253.5 KB 9b56c08f4b07
samples/006/003.wavOther357.2 KB 61f0c9abade2
samples/006/004.wavOther157.5 KB 5dce004d4422
samples/007/000.wavOther326.4 KB dfd4df27aff6
samples/007/001.wavOther314.9 KB a740777bf1a1
samples/007/002.wavOther230.4 KB 6a6735e825e8
samples/007/003.wavOther372.5 KB 48af1ec2b374
samples/007/004.wavOther161.3 KB 2e84f57b2b02
samples/008/000.wavOther334.1 KB c4416182b93a
samples/008/001.wavOther303.4 KB 42cf44260f3d
samples/008/002.wavOther242.0 KB 58481824c14c
samples/008/003.wavOther407.1 KB 04e40e47d57e
samples/008/004.wavOther153.6 KB d81155a18713
samples/009/000.wavOther307.2 KB 7027d49750ab
samples/009/001.wavOther299.6 KB 1b4b5703d2bf
samples/009/002.wavOther203.6 KB 15620cc54c65
samples/009/003.wavOther387.9 KB e24a8df8ed38
samples/009/004.wavOther138.3 KB 02082f5b18ce
samples/010/000.wavOther399.4 KB 0c647a71b1c5
samples/010/001.wavOther318.8 KB 2fc2f2f09043
samples/010/002.wavOther242.0 KB 64500f2f305c
samples/010/003.wavOther364.8 KB 6682c6bc4b02
samples/010/004.wavOther122.9 KB e86dbf361cee
samples/011/000.wavOther357.2 KB e7b26b394337
samples/011/001.wavOther334.1 KB f3c5f3105459
samples/011/002.wavOther242.0 KB f59694c48e39
samples/011/003.wavOther384.0 KB 753169255a30
samples/011/004.wavOther161.3 KB a00aa678d14e
samples/012/000.wavOther368.7 KB 0846fd92d26d
samples/012/001.wavOther422.4 KB e4b0ee206895
samples/012/002.wavOther249.6 KB e42bf4e4e1c1
samples/012/003.wavOther357.2 KB 2e640e26105a
samples/012/004.wavOther134.4 KB 6bbb98dc0478
samples/013/000.wavOther334.1 KB 6f4acff2c822
samples/013/001.wavOther307.2 KB 5645efcc6bce
samples/013/002.wavOther242.0 KB 2b8791455167
samples/013/003.wavOther361.0 KB 89c1140b9d8c
samples/013/004.wavOther138.3 KB 12ab72cf6e70
samples/014/000.wavOther364.8 KB 6e17f64bdee6
samples/014/001.wavOther353.3 KB 4d5c00df97e6
samples/014/002.wavOther268.8 KB 67ab3a5c06dc
samples/014/003.wavOther391.7 KB 843a91588704
samples/014/004.wavOther142.1 KB 87a2527990a1
training-checkpoint/latestOther13 B —
training-checkpoint/random_states_0.pklOther15.1 KB 7d20cee112ae
training-checkpoint/random_states_1.pklOther15.0 KB 342d3108d2b0
.gitattributesRepository5.7 KB —
merges.txtTokenizer1.7 MB —
tokenizer_config.jsonTokenizer7.3 KB —
vocab.jsonTokenizer2.8 MB —

License and Download

License
Not stated by the source
Access
Open weights, no gate
Download size
6.5 GB
Download from Rijul Paul

Released by Rijul Paul through its official repository on Hugging Face.

Memory Requirements

PrecisionWeights in memory
As published6.5 GB
16-bit1.8 GB
8-bit0.9 GB
4-bit0.5 GB

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

Questions About Qwen3-TTS-12Hz-0.6B-Zhongli-LoRa-32

How much GPU memory does Qwen3-TTS-12Hz-0.6B-Zhongli-LoRa-32 need?

About 2.2 GB at 16-bit and 0.5 GB at 4-bit: the weights (906M parameters) plus a working margin. A long context needs more.

What is the cheapest GPU to run Qwen3-TTS-12Hz-0.6B-Zhongli-LoRa-32 on?

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