K2-Horizon-7B-oQ3e is an open-weight model from MLX Works. It has 9B parameters and a 524,288-token context. At 16-bit it needs about 21.6 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index.
This model was quantized using oQ (oMLX v0.7.0) mixed-precision quantization.
Runs On
What it takes to serve K2-Horizon-7B-oQ3e (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 | 18.0 GB | 21.6 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 9.0 GB | 10.8 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 4.5 GB | 5.4 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 23, 2026.
K2-Horizon-7B-oQ3e on every accelerator the SAVRN Index prices, at every precision
Model Card
This model was quantized using oQ (oMLX v0.7.0) mixed-precision quantization.
Excerpt from the card by MLX Works.
Configuration
- Architecture
- K2HorizonForCausalLM
- Context length (tokens)
- 524,288
- Layers
- 36
- Hidden size
- 4,096
- Feed-forward size
- 12,288
- Attention heads
- 32
- Key/value heads
- 8
- Head dimension
- 128
- Vocabulary size
- 250,624
- Experts
- 0
- Experts active per token
- 0
- Model type
- k2_horizon
Identity and Version
- Repository
- mlx-works/K2-Horizon-7B-oQ3e
- Publisher
- MLX Works
- Task
- Not stated by the source
- Modality
- Other
- Library
- mlx
- Parameters
- 9B parameters
- Languages
- mlx, oq
- Revision
- 232b352155370d51abdaf63134949899b19ffae5
- First published
- 2026-10-01
- Last updated
- 2026-10-01
Files and Weights
11 files, 4.2 GB in total. The weights are 1 file totalling 4.2 GB in safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model.safetensors | Weights | 4.2 GB | 32337ea30fa4 |
| config.json | Configuration | 9.0 KB | — |
| configuration_k2_horizon.py | Configuration | 3.5 KB | — |
| generation_config.json | Configuration | 81 B | — |
| modeling_k2_horizon.py | Configuration | 49.5 KB | — |
| oq_imatrix_report.json | Configuration | 12.7 KB | — |
| README.md | Documentation | 310 B | — |
| chat_template.jinja | Other | 51.0 KB | — |
| .gitattributes | Repository | 1.6 KB | — |
| tokenizer.json | Tokenizer | 20.6 MB | 838d767b7c99 |
| tokenizer_config.json | Tokenizer | 293 B | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 4.2 GB
Released by MLX Works through its official repository on Hugging Face.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 4.2 GB |
| 16-bit | 18.0 GB |
| 8-bit | 9.0 GB |
| 4-bit | 4.5 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About K2-Horizon-7B-oQ3e
How much GPU memory does K2-Horizon-7B-oQ3e need?
About 21.6 GB at 16-bit and 5.4 GB at 4-bit: the weights (9B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run K2-Horizon-7B-oQ3e 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.
What is K2-Horizon-7B-oQ3e's context length?
524,288 tokens, from the maximum position embeddings in its published configuration.