AI Potluck
Model components / Base / pretrained models

Falcon 3

Technology Innovation Institute

TII's open-weights model line from Abu Dhabi. The Falcon 3 family spans 1B to 10B and was trained on 14 trillion tokens of web, code and STEM data. It has since been extended by Falcon-H1, a hybrid Mamba-transformer architecture from 0.5B to 34B with native support for eighteen languages, and by an Arabic-specialized Falcon-H1 line at 3B, 7B and 34B.

Scored on the Falcon 3 generation, now dated relative to later families. The Hub does not name the license, so the tier rests on TII's own published text. Adoption was re-banded on a measured download count rather than reported traction. Verified 2026-08-13 via the Falcon 3 release blog and the Falcon3 model cards.

Openness

3 high confidence
3.0
weights
open
data
closed
code
closed
license
TII-Falcon-License-2.0(Apache-based+AUP,non-OSI)

Weights are downloadable and freely usable commercially at any scale, but the corpus is not released and there is no training code, so 3. Not 2: the TII-Falcon-License-2.0 is Apache-based with an acceptable-use policy, and a conduct restriction caps neither who may use the model nor at what scale, so it belongs with permissive non-OSI licenses and attribution requirements rather than with capped ones such as Llama's 700M-MAU ceiling. Not 4 or 5 either, since corpus and pipeline are both closed and the license is non-OSI in any case - the open-source definition forbids discriminating against a field of endeavor. Hugging Face reports the license as 'other', an abstention rather than an answer, so the reading rests on TII's published text.

Adoption

1 high confidence
1.0

The single declared artifact tiiuae/Falcon3-10B-Base returns 5,292 downloads in the trailing 30 days, alongside 41 likes, which bands at level 1 (<10K) on the model adoption scale. A measured download count outranks a reported-traction read, and here it confirms what the family's own description already said: relevance has faded against newer small language models.

Capability

2 high confidence
2.0

Falcon3-10B-Base: MMLU 73.1, MMLU-Pro 42.5, GSM8K 83.0; Falcon3-7B-Base: MMLU 67.4, MMLU-Pro 39.2. Competitive among sub-13B in 2024 but low vs 2026 frontier.

Verified 2026-08-13