AlphaBet_memory-on is an open-weight model from Liqin Ye. It has 106.9B parameters and a 131,072-token context. At 16-bit it needs about 256.4 GB of GPU memory, which fits on 1x MI355X from $2.59 an hour; at 4-bit, 64.1 GB on 1x MI300X from $1.85, at the lowest prices in the SAVRN Index.
Runs On
What it takes to serve AlphaBet_memory-on (106.9B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
|---|---|---|---|---|---|
| 16-bit | 213.7 GB | 256.4 GB | 1x MI355X (288 GB) Vultr |
$2.59 | 2x MI300X $3.70 · 2x MI325X $4.00 |
| 8-bit | 106.9 GB | 128.2 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x MI325X $2.00 · 1x MI355X $2.59 |
| 4-bit | 53.4 GB | 64.1 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 24, 2026.
AlphaBet_memory-on on every accelerator the SAVRN Index prices, at every precision
Model Card
The publisher has not written a card for this model.
Configuration
- Architecture
- Glm4MoeForCausalLM
- Context length (tokens)
- 131,072
- Layers
- 46
- Hidden size
- 4,096
- Feed-forward size
- 10,944
- Attention heads
- 96
- Key/value heads
- 8
- Head dimension
- 128
- Vocabulary size
- 151,552
- Routed experts
- 128
- Experts active per token
- 8
- RoPE base
- 1,000,000
- Stored precision
- bfloat16
- Model type
- glm4_moe
Identity and Version
- Repository
- liqinye/AlphaBet_memory-on
- Publisher
- Liqin Ye
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- 106.9B parameters
- Languages
- Not stated by the source
- Revision
- c0cbc2ee3ac864e25a42253ad20d921bb17088cd
- First published
- 2026-10-11
- Last updated
- 2026-10-11
Files and Weights
47 files, 213.7 GB in total. The weights are 40 files totalling 213.7 GB in safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model-00000-of-00040.safetensors | Weights | 5.4 GB | 622c07521362 |
| model-00001-of-00040.safetensors | Weights | 5.4 GB | 24b55a0dbdb6 |
| model-00002-of-00040.safetensors | Weights | 5.4 GB | 29f8aab9594e |
| model-00003-of-00040.safetensors | Weights | 5.4 GB | e59dec8c7f13 |
| model-00004-of-00040.safetensors | Weights | 5.4 GB | cf8a8cd9b240 |
| model-00005-of-00040.safetensors | Weights | 5.4 GB | 319ff2d46c12 |
| model-00006-of-00040.safetensors | Weights | 5.4 GB | 789f61426019 |
| model-00007-of-00040.safetensors | Weights | 5.4 GB | 1b11b4d5d0c3 |
| model-00008-of-00040.safetensors | Weights | 5.4 GB | 9f198777da07 |
| model-00009-of-00040.safetensors | Weights | 5.4 GB | 5dc6367d27dc |
| model-00010-of-00040.safetensors | Weights | 5.4 GB | 10b94b508887 |
| model-00011-of-00040.safetensors | Weights | 5.4 GB | 45fc5b54df5e |
| model-00012-of-00040.safetensors | Weights | 5.4 GB | baea52fab378 |
| model-00013-of-00040.safetensors | Weights | 5.4 GB | e4b6993c0caa |
| model-00014-of-00040.safetensors | Weights | 5.4 GB | 690bcb04cdd8 |
| model-00015-of-00040.safetensors | Weights | 5.4 GB | 55f506ae4281 |
| model-00016-of-00040.safetensors | Weights | 5.4 GB | 3675ba1f353a |
| model-00017-of-00040.safetensors | Weights | 5.4 GB | 85822b9f752d |
| model-00018-of-00040.safetensors | Weights | 5.4 GB | 142683980566 |
| model-00019-of-00040.safetensors | Weights | 5.4 GB | 15c848f47efe |
| model-00020-of-00040.safetensors | Weights | 5.4 GB | 9c299cf1317a |
| model-00021-of-00040.safetensors | Weights | 5.4 GB | dedb726c4ebd |
| model-00022-of-00040.safetensors | Weights | 5.4 GB | ec4cc1ccfcf7 |
| model-00023-of-00040.safetensors | Weights | 5.4 GB | 0bfbfef01f14 |
| model-00024-of-00040.safetensors | Weights | 5.4 GB | 1eabceea6835 |
| model-00025-of-00040.safetensors | Weights | 5.4 GB | 5df01bdb567b |
| model-00026-of-00040.safetensors | Weights | 5.4 GB | 9b015c6f2d8a |
| model-00027-of-00040.safetensors | Weights | 5.4 GB | 5e64d7068b0e |
| model-00028-of-00040.safetensors | Weights | 5.4 GB | 7bdaa499a240 |
| model-00029-of-00040.safetensors | Weights | 5.4 GB | c9569a03667b |
| model-00030-of-00040.safetensors | Weights | 5.4 GB | db59de9cd77a |
| model-00031-of-00040.safetensors | Weights | 5.4 GB | 8969be4abc89 |
| model-00032-of-00040.safetensors | Weights | 5.4 GB | c7163f1fcebf |
| model-00033-of-00040.safetensors | Weights | 5.4 GB | 246404ccfc9c |
| model-00034-of-00040.safetensors | Weights | 5.4 GB | 2d1a8510cd2c |
| model-00035-of-00040.safetensors | Weights | 5.4 GB | 381a836700d7 |
| model-00036-of-00040.safetensors | Weights | 5.4 GB | da34d3df999b |
| model-00037-of-00040.safetensors | Weights | 5.4 GB | aa473429ecea |
| model-00038-of-00040.safetensors | Weights | 5.4 GB | 0ee3ae6a6574 |
| model-00039-of-00040.safetensors | Weights | 4.5 GB | b8ddb011f410 |
| config.json | Configuration | 1.0 KB | — |
| generation_config.json | Configuration | 155 B | — |
| model.safetensors.index.json | Configuration | 1.6 MB | — |
| chat_template.jinja | Other | 3.2 KB | — |
| .gitattributes | Repository | 1.6 KB | — |
| tokenizer.json | Tokenizer | 20.0 MB | 934066501641 |
| tokenizer_config.json | Tokenizer | 7.3 KB | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 213.7 GB
Released by Liqin Ye through its official repository on Hugging Face.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 213.7 GB |
| 16-bit | 213.7 GB |
| 8-bit | 106.9 GB |
| 4-bit | 53.4 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About AlphaBet_memory-on
How much GPU memory does AlphaBet_memory-on need?
About 256.4 GB at 16-bit and 64.1 GB at 4-bit: the weights (106.9B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run AlphaBet_memory-on 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.
What is AlphaBet_memory-on's context length?
131,072 tokens, from the maximum position embeddings in its published configuration.