Model components / Base / pretrained models

SmolLM3

Hugging Face's fully-open 3B small-model line (base + Instruct), released Jul 2025. Trained on ~11T tokens with a 128K context window and 6-language multilingual support; dual-mode reasoning. The full engineering blueprint (architecture, data mixtures, training and post-training recipe) is published. Supersedes SmolLM2.

Apache-2.0 weights; public data mixtures and training recipe. Outperforms Llama-3.2-3B and Qwen2.5-3B and stays competitive with 4B Qwen3/Gemma3; light enough to run on-device.

Openness

5 high confidence

5.0

weights: open(Apache-2.0)
data: open(public data mixtures, ~11T tokens)
code: open(training+post-training recipe published)
license: Apache-2.0(OSI)

Fully-open small-model line: Apache-2.0 weights with the complete engineering blueprint (architecture, data mixtures, training and post-training recipe) published.

https://huggingface.co/blog/smollm3 recorded 2026-06-30
fully-open 3B: 11T tokens, 128K context, multilingual, full recipe published
https://huggingface.co/HuggingFaceTB/SmolLM3-3B recorded 2026-06-30
SmolLM3-3B on HF; Apache-2.0

Adoption

3 medium confidence

3.0

Popular fully-open small model (HF's SmolLM line saw ~184k monthly downloads at the SmolLM2 generation); strong in on-device/edge use, niche vs the large open-weight families.

https://huggingface.co/HuggingFaceTB/SmolLM3-3B recorded 2026-06-30
active SmolLM3-3B HF repo

Capability

3 high confidence

3.0

Strong intelligence-per-parameter for a 3B model, but small absolute capability vs larger open and open-weight models.

https://huggingface.co/blog/smollm3 recorded 2026-06-30
beats Llama-3.2-3B and Qwen2.5-3B; competitive with 4B alternatives

Unchanged since 2026-07-30 (last edited, not re-checked)