LLM EXPLORER 66,431 MODELS INDEXED

Atomight V2.5 1.7B by NovatasticRoScript

By NovatasticRoScript · 1406 downloads
HF Repo ↗ Copy URL

Atomight V2.5 1.7B is an open-source language model by NovatasticRoScript. Features: 1.7b LLM, VRAM: 3.4GB, Context: 40K, License: cc-by-4.0, LLM Explorer Score: 0.27.

  Arxiv:2402.03300   Arxiv:2505.09388   Atomight   Atomightv2-5 Base model:finetune:qwen/qwen3...   Base model:qwen/qwen3-1.7b   Conversational Dataset:logics-mllm/logics-ste... Dataset:megascience/megascienc... Dataset:nvidia/nemotron-sft-co... Dataset:nvidia/opencodeinstruc... Dataset:nvidia/openmathreasoni...   Dataset:trl-lib/deepmath-103k   Dataset:zwhe99/deepmath-103k   En   Eval-results   Grpo   Qwen3   Reasoning   Region:us   Safetensors
Model Card on HF 🤗: https://huggingface.co/NovatasticRoScript/Atomight-V2.5-1.7B 

Atomight V2.5 1.7B Parameters and Internals

LLM NameAtomight V2.5 1.7B
Repository 🤗https://huggingface.co/NovatasticRoScript/Atomight-V2.5-1.7B 
Base Model(s)  Qwen/Qwen3-1.7B   Qwen/Qwen3-1.7B
Model Size1.7b
Required VRAM3.4 GB
Updated2026-08-06
MaintainerNovatasticRoScript
Model Typeqwen3
Model Files  3.4 GB
Supported Languagesen
Model ArchitectureQwen3ForCausalLM
Licensecc-by-4.0
Context Length40960
Model Max Length40960
Tokenizer ClassQwen2Tokenizer
Padding Token<|vision_pad|>
Vocabulary Size151936
Torch Data Typefloat16
Errorsreplace

Best Alternatives to Atomight V2.5 1.7B

Best Alternatives
Context / RAM
Downloads
Likes
Reaper Base128K / 3.4 GB3930
Lucy 128K128K / 3.4 GB123112
Polaris 1.7B Preview128K / 3.4 GB738
DictaLM 3.0 1.7B Instruct60K / 3.4 GB28421
Qwen3 1.7B40K / 4 GB5633291500
Boole 1.7B40K / 0.2 GB5399
Qwen3 1.7B Openvino40K / 0 GB1571
Qwen3 1.7B Sudoku SFT40K / 3.4 GB2570
Terrible Joke Qwen3 1.7B40K / 3.4 GB1090
Stormdesks Triage40K / 3.4 GB3091
Note: green Score (e.g. "73.2") means that the model is better than NovatasticRoScript/Atomight-V2.5-1.7B.