KeyLM-75M
KeyLM-75M is a 75M parameter base language model trained from scratch on approximately 18 billion tokens. That training budget is a small fraction of what comparable small models use (SmolLM-135M was trained on roughly 600B tokens, SmolLM2-135M on roughly 2T).
This is the
base model: a text-completion model, not instruction-tuned. For chat and instruction following, use
KeyLM-75M-Instruct.
Table of Contents
- Model Summary
- How to Use
- Evaluation
- Training
- Limitations
- License
- Citation
Model Summary
KeyLM is a compact decoder-only transformer built on the standard small-model recipe used by Llama and Qwen3: grouped-query attention, rotary position embeddings (RoPE), SwiGLU feed-forward layers, and per-head QK-RMSNorm.
| Field | Value |
|---|
| Parameters | 75,251,200 |
| Layers | 24 |
| Hidden size | 512 |
| Attention heads | 8 (2 KV heads, GQA) |
| Context length | 2048 |
| Vocabulary | 12,020 (ByteLevel BPE) |
| Precision | bfloat16 |
| Training tokens | ~18B |
How to Use
1import torch
2from transformers import AutoModelForCausalLM, AutoTokenizer
3
4model_id = "Eclipse-Senpai/KeyLM-75M"
5tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
6model = AutoModelForCausalLM.from_pretrained(
7 model_id, trust_remote_code=True, torch_dtype=torch.bfloat16
8)
9
10inputs = tokenizer("The three primary colors are", return_tensors="pt")
11outputs = model.generate(
12 **inputs, max_new_tokens=40, do_sample=True,
13 temperature=0.7, top_p=0.9, repetition_penalty=1.1,
14)
15print(tokenizer.decode(outputs[0], skip_special_tokens=True))
Evaluation
On zero-shot benchmarks (lm_eval; accuracy, with length-normalized accuracy for ARC and HellaSwag) KeyLM is modest but above random on basic commonsense, and at chance on knowledge-heavy tasks.
| Benchmark | KeyLM-75M (base) | KeyLM-75M-Instruct | Random |
|---|
| IFEval (4-metric avg) | — | 17.85 | — |
| MMLU | 23.0 | 24.0 | 25.0 |
| ARC (avg) | 29.9 | 30.8 | 25.0 |
| HellaSwag | 29.7 | 31.0 | 25.0 |
| PIQA | 60.0 | 61.3 | 50.0 |
| WinoGrande | 48.4 | 48.3 | 50.0 |
| OpenBookQA | 25.0 | 25.0 | 25.0 |
Instruction tuning leaves knowledge and reasoning roughly unchanged, its effect is the instruction-following ability (IFEval) the base lacks.
Training
KeyLM-75M was pretrained from random initialization on approximately 18B tokens, drawn from a weighted mixture of public datasets streamed through a deterministic curriculum.
| Category | Share | Sources |
|---|
| Formal / quality | ~30% | FineWeb-Edu, Wikipedia |
| Casual / social | ~30% | Reddit comments, StackExchange |
| Conversational | ~25% | WildChat, UltraChat, LMSYS-Chat, OASST2 |
| Structured knowledge | ~5% | Cosmopedia |
| Typo augmentation | ~10% | Synthetic (contrastive) |
The instruction-tuned model built on this base is available at
KeyLM-75M-Instruct.
Limitations
- Minimal world knowledge. Not suitable for factual question answering, reasoning, math, or code.
- Base model: it completes text and does not follow instructions or hold a conversation. Use the Instruct version for chat.
- English only.
- No safety alignment. Apply your own filtering before any user-facing use.
License
Apache 2.0. The weights are trained from scratch and free to use, modify, and redistribute.
Citation
1@misc{keylm75m2026,
2 title = {KeyLM-75M: a from-scratch small language model},
3 author = {Eclipse-Senpai},
4 year = {2026},
5 howpublished = {\url{https://huggingface.co/Eclipse-Senpai/KeyLM-75M}}
6}
