Atomight-V2.5-1.7B

<div align="center"> <img src="NewAtomight.png" alt="Atomight v2 Logo" width="500" style="max-width: 100%;">

Atomight V2.5 · 1.7B

Reasoning-first · Zero benchmark contamination · Trained on a free Colab T4

</div>

---

We are excited to announce and show you all, our most powerful and capable model in the current and newest Atomight family variant (V2.5).

[!Note] Use this model when you want explicit chain-of-thought before the final answer — complex debugging, multi-step planning, agentic workflows, and math- or reasoning-heavy tasks.

[!Note] Knowledge Span/Cutoff: Late 2024 / Early 2025 – Late 2025

Other details for this model soon. Wait for further information and details.

---

Quick Facts

---

Training Data

Curated (not scraped) from frontier 2025-era open datasets across STEM, science, math, and code:

[!Important] No benchmark data was used in training. None of the sources above overlap with MMLU, GSM8K, HumanEval, MBPP, HellaSwag, WinoGrande, ARC-Challenge, or TruthfulQA. Every score below was earned, not leaked.

---

Benchmark results

These are the results of the benchmarks for Atomight-V2.5-1.7B, evaluated with lm-evaluation-harness.

_{All scores self-reported; no training data overlaps with these benchmarks — see Training Data section above.}

---

IIfSLM Benchmark Results

IIfSLM (Intelligence Index for Small Language Models) is an open, contamination-resistant benchmark suite for the 0.5B–3B range — built for the whole small-model community to evaluate against, not exclusive to this model.

Methodology and full results are documented in the accompanying paper:

NovatasticRoScript (2026). IIfSLM: An Intelligence Index for Small Language Models. Zenodo. https://doi.org/10.5281/zenodo.21753925

See the full IIfSLM dataset and methodology notes -- and feel free to run your own model against it too.

---

How it compares with other small language models (we recommend verifying it, as the other data from other models came from a third-party sources)

Scores below for other models are drawn from their respective model cards / technical reports, not re-run by us. Provided for context only — evaluation harnesses and prompt formats differ across labs, so treat this as directional rather than exact.

Where Atomight-V2.5-1.7B leads: best-in-class GSM8K among comparable 1–2B base models, and MMLU well above SmolLM2-1.7B — achieved with a curated, contamination-free ~2K-per-category dataset rather than large-scale pretraining.

Where it trails: HumanEval and MBPP lag behind instruction-tuned peers like InfiR-1B-Instruct and Qwen2.5-1.5B-Instruct — expected, since Atomight-V2.5-1.7B is a base reasoning model without dedicated code SFT. HellaSwag also trails models trained on broader web/commonsense data.

---

Quick Start

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "NovatasticRoScript/Atomight-V2.5-1.7B"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")

messages = [
    {"role": "system", "content": "You are a reasoning model. Think step-by-step inside <thinking> tags, then give your final answer inside <answer> tags."},
    {"role": "user", "content": "If a train travels 60 miles in 45 minutes, what is its speed in mph?"}
]

inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)
outputs = model.generate(inputs, max_new_tokens=512)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Main citation

This model is a fine-tuned derivative of Qwen3-1.7B (Apache 2.0) and is released under CC-BY-4.0 in accordance with the licenses of its training datasets. If you use this model, kindly credit NovatasticRoScript/Atomight-V2.5-1.7B and the base model/datasets listed above. Part of the Atomight family — small models, curated data, no shortcuts.

Other citacion/s

@misc{qwen3technicalreport,
      title={Qwen3 Technical Report}, 
      author={Qwen Team},
      year={2025},
      eprint={2505.09388},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2505.09388}, 
}

Training method — this model was trained using GRPO (Group Relative Policy Optimization), introduced in:

@article{deepseekmath2024,
  title={DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models},
  author={Shao, Zhihong and Wang, Peiyi and Zhu, Qihao and Xu, Runxin and Song, Junxiao and Zhang, Mingchuan and Li, Y.K. and Wu, Y. and Guo, Daya},
  journal={arXiv preprint arXiv:2402.03300},
  year={2024}
}

@misc{iifslm2026,
  author       = {NovatasticRoScript},
  title        = {IIfSLM: An Intelligence Index for Small Language Models — A Contamination-Resistant, Community-Driven Benchmark Suite for the 0.5B–3B Parameter Range},
  year         = {2026},
  publisher    = {Zenodo},
  doi          = {10.5281/zenodo.21753925},
  url          = {https://doi.org/10.5281/zenodo.21753925}
}