This model is a fine-tuned version of Qwen/Qwen3.8-27B on the on the Omni-Edu-70K dataset. The following hyperparameters were used during training: - learningrate: 5e-06 - trainbatchsize: 1 - evalbatchsize: 8 - distributedtype: multi-GPU - numdevices: 16 - gradientaccumulationsteps: 8 - totaltrainbatchsize: 128 - totalevalbatchsize: 128 - lrschedulertype: cosine - lrschedulerwarmupsteps: 0.1 - numepochs: 3.0 - Transformers 5.2.0 - Pytorch 2.10.0 - Datasets 4.0.0 - Tokenizers 0.22.2
Open-weight model · Image and text to text
TerminalHorizon-27B
by Zoushun shunzou05/TerminalHorizon-27B
TerminalHorizon-27B is an open-weight model for image and text to text from Zoushun, released under Apache License 2.0. It has 3M parameters and a 262,144-token context. At 16-bit it needs about 0 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index. It draws 27 downloads a month.
Scaling Agentic Data for Long-Horizon Terminal Intelligence TerminalHorizon is a fully automated data synthesis engine for long-horizon terminal agents.
Parameters3M
Context262,144
Weights54.7 GB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads27
Runs On
What it takes to serve TerminalHorizon-27B (3M 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 | 0.0 GB | 0.0 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 0.0 GB | 0.0 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 0.0 GB | 0.0 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 18, 2026.
TerminalHorizon-27B on every accelerator the SAVRN Index prices, at every precision
Model Card
By Zoushun, published under apache-2.0, revision 382656f498b3.
Scaling Agentic Data for Long-Horizon Terminal Intelligence TerminalHorizon is a fully automated data synthesis engine for long-horizon terminal agents. Grounded in real-world professional work, it constructs executable environments and generates training trajectories that connect agentic behavioral patterns across stages toward a shared goal. This repository releases TerminalHorizon-27B, obtained by supervised fine-tuning Qwen3.5-27B on The terminal is a shared interaction interface, not a restriction to software tasks. The training environments span 25 professional domains and 463 subdomains, including software engineering, machine learning, education, agriculture, healthcare, finance…
Read Zoushun's full model card
Scaling Agentic Data for Long-Horizon Terminal Intelligence
Paper · Citation
TerminalHorizon is a fully automated data synthesis engine for long-horizon terminal agents. Grounded in real-world professional work, it constructs executable environments and generates training trajectories that connect agentic behavioral patterns across stages toward a shared goal.
This repository releases TerminalHorizon-27B, obtained by supervised fine-tuning Qwen3.5-27B on TerminalHorizon-3K.
Resources
| Resource | Description |
|---|---|
| TerminalHorizon-Environment | Executable terminal-task environments. |
| TerminalHorizon-3K | Long-horizon training trajectories. |
| TerminalHorizon-27B | Fine-tuned Qwen3.5-27B. |
| TerminalHorizon-35B-A3B | Fine-tuned Qwen3.5-35B-A3B. |
| TerminalHorizon-122B-A10B | Fine-tuned Qwen3.5-122B-A10B. |
Model Overview
| Property | Value |
|---|---|
| Base model | Qwen/Qwen3.5-27B |
| Training method | Supervised fine-tuning |
| Training dataset | TerminalHorizon-3K |
| Training trajectories | 2,852 |
| Weight precision | BF16 |
| Checkpoint format | Hugging Face Transformers / Safetensors |
The terminal is a shared interaction interface, not a restriction to software tasks. The training environments span 25 professional domains and 463 subdomains, including software engineering, machine learning, education, agriculture, healthcare, finance, and scientific research.
TerminalHorizon focuses on meaningful work and connected agent behavior, rather than task length alone.
Load the Model
Use a version of Transformers that supports Qwen3.5:
from transformers import AutoProcessor, Qwen3_5ForConditionalGeneration
model_id = "shunzou05/TerminalHorizon-27B"
processor = AutoProcessor.from_pretrained(model_id)
model = Qwen3_5ForConditionalGeneration.from_pretrained(
model_id,
dtype="auto",
device_map="auto",
)
Refer to the base model documentation for inference and serving examples.
Citation
@misc{zou2026scalingagenticdatalonghorizon,
title={Scaling Agentic Data for Long-Horizon Terminal Intelligence},
author={Shun Zou and Ziyu Ma and Yi Zou and Yong Wang and Lin Chen and Zehui Chen and Guanhua Chen and Xiangxiang Chu and Feng Zhao},
year={2026},
eprint={2610.13166},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2610.13166},
}
Configuration
- Architecture
- Qwen3_5ForConditionalGeneration
- Context length (tokens)
- 262,144
- Layers
- 64
- Hidden size
- 5,120
- Feed-forward size
- 17,408
- Attention heads
- 24
- Key/value heads
- 4
- Head dimension
- 256
- Vocabulary size
- 248,320
- Model type
- qwen3_5
Identity and Version
- Repository
- shunzou05/TerminalHorizon-27B
- Publisher
- Zoushun
- Task
- Image and text to text
- Modality
- Image and text
- Library
- transformers
- Parameters
- 3M parameters
- Languages
- en
- Revision
- 382656f498b3deb367debf82165d2d2a7132e8c9
- First published
- 2026-10-06
- Last updated
- 2026-10-12
Files and Weights
16 files, 54.7 GB in total. The weights are 2 files totalling 54.7 GB in safetensors.
Weights2 files · 54.7 GB
Configuration6 files · 117.1 KB
Tokenizer4 files · 28.6 MB
Documentation1 file · 3.4 KB
Other2 files · 228.3 KB
Repository1 file · 1.6 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model-00001-of-00002.safetensors | Weights | 49.8 GB | 669c23b9b35f |
| model-00002-of-00002.safetensors | Weights | 4.9 GB | 3187ff6c2df2 |
| config.json | Configuration | 3.8 KB | — |
| generation_config.json | Configuration | 199 B | — |
| model.safetensors.index.json | Configuration | 111.1 KB | — |
| preprocessor_config.json | Configuration | 390 B | — |
| processor_config.json | Configuration | 1.3 KB | — |
| video_preprocessor_config.json | Configuration | 385 B | — |
| README.md | Documentation | 3.4 KB | — |
| assets/data-analysis.webp | Other | 220.6 KB | bc54ebd8ca57 |
| chat_template.jinja | Other | 7.8 KB | — |
| .gitattributes | Repository | 1.6 KB | — |
| merges.txt | Tokenizer | 3.4 MB | — |
| tokenizer.json | Tokenizer | 20.0 MB | 87a7830d63fc |
| tokenizer_config.json | Tokenizer | 1.2 KB | — |
| vocab.json | Tokenizer | 5.2 MB | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 54.7 GB
Download from Zoushun
Released by Zoushun through its official repository on Hugging Face. Read the license.
Built From
- Derived from Qwen/Qwen3.5-27B
- Described by arXiv:2610.13166
- Trained on (disclosed) shunzou05/TerminalHorizon-3K
- Trained on (disclosed) shunzou05/TerminalHorizon-Environment
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 54.7 GB |
| 16-bit | 0.0 GB |
| 8-bit | 0.0 GB |
| 4-bit | 0.0 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About TerminalHorizon-27B
How much GPU memory does TerminalHorizon-27B need?
About 0 GB at 16-bit and 0 GB at 4-bit: the weights (3M parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run TerminalHorizon-27B 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.
Can I use TerminalHorizon-27B commercially?
Yes. TerminalHorizon-27B is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.
What is TerminalHorizon-27B's context length?
262,144 tokens, from the maximum position embeddings in its published configuration.
Similar Models
Full fine-tune of Qwen/Qwen3-VL-8B-Instruct (revision 0c351dd) for the box-conditioned variant of the deleafing cut-point task: the input is one robot head-camera frame (848x408 RGB) with its aligned depth map (metres, fed as a rendered second image) and the bounding box of the target petiole; the output is the nominal cut point 9 mm along that petiole from its junction with the main stem. The model does not choose the target. Compared with the box+point models in this account (qwen3-vl-8b-tomato-cutpoint-bp-, which find the petiole themselves), this one measures localization given identity. Answer: {"cutpointuv":[x,y]} in normalized [0,1000) coordinates of the original 848x408 frame (2…
This model is a fine-tuned version of /mnt/shared-storage-user/mineru2-shared/niujunbo/ldy/models/Qwen3.5-35B-A3B-Instruct on the merged8083ulogocohqwen35 dataset. The following hyperparameters were used during training: - learningrate: 5e-05 - trainbatchsize: 1 - evalbatchsize: 8 - distributedtype: multi-GPU - numdevices: 8 - gradientaccumulationsteps: 8 - totaltrainbatchsize: 64 - totalevalbatchsize: 64 - lrschedulertype: cosine - lrschedulerwarmupsteps: 0.1 - numepochs: 1.0 - Transformers 5.6.0 - Pytorch 2.10.0+cu128 - Datasets 4.0.0 - Tokenizers 0.22.2
Scaling Agentic Data for Long-Horizon Terminal Intelligence TerminalHorizon is a fully automated data synthesis engine for long-horizon terminal agents. Grounded in real-world professional work, it constructs executable environments and generates training trajectories that connect agentic behavioral patterns across stages toward a shared goal. This repository releases TerminalHorizon-35B-A3B, obtained by supervised fine-tuning Qwen3.5-35B-A3B on The terminal is a shared interaction interface, not a restriction to software tasks. The training environments span 25 professional domains and 463 subdomains, including software engineering, machine learning, education, agriculture, healthcare…
Model · Image and text to text
Qwen3.8-Flash-Next-GSQ-RCO-GGUF
Non-uniform GGUF quantizations of a 512-expert MoE, produced with GSQ and RCO, with a vision projector for multimodal use. This repository provides GGUF quantizations of Qwen3.8-Flash-Next at four sizes, together with the model's vision projector (mmproj) for multimodal use. In contrast to uniform quantization, which applies a single quantization type to all weight tensors, each model here assigns a separate quantization type to every tensor. The assignment is obtained by a gradient-based search that allocates precision according to per-tensor sensitivity, subject to a total size budget. The resulting files are standard GGUF and run unmodified in llama.cpp, Ollama, and LM Studio. A…
I built this quant because the ready-made FP4 file answered the wrong question. It was fast, but on my short WikiText-2 control it scored 6.4949 PPL. Plain Q40 scored 6.3798. The first higher-quality hybrid went too far the other way: good perplexity, 34.19 tok/s, and no comfortable room for 256K plus vision. This is the build that survived both gates. It is a 17.1 GB, 5.01 BPW mixed-precision GGUF of Qwen/Qwen3.8-27B. It keeps large, tolerant matrices in native NVFP4 and spends more bits on selected attention, Gated DeltaNet, and late FFN tensors. The trained MTP layer remains embedded in the same GGUF. This is not a fine-tune. I built the private calibration workload from 5,472 messages…