OpenELM

<style>
.ios-ib th.ios-lab { background-color: #333; color: #fff; text-align: center; width: 28%; padding: 6px; }
.ios-ib td.ios-val { background-color: #fff; text-align: left; padding: 6px 10px; }
@dark { .ios-ib td.ios-val { background-color: var(--wiki-surface); } }
</style>
{| class="ios-ib" style="width: 100%; border: 2px solid #333; text-align: center;" darkstyle="border-color: #48484a;"
|-
| colspan="2" style="padding: 28px 20px 42px;" |
<div style="font: bold 2em/1.2 -apple-system, BlinkMacSystemFont, 'SF Pro Display', 'Helvetica Neue', Helvetica, Arial, sans-serif; color: #333;" darkstyle="color: var(--wiki-color);">OpenELM</div>
|-
! class="ios-lab" | Developer
| class="ios-val" | [[Apple]] machine learning research
|-
! class="ios-lab" | Released
| class="ios-val" | April 2024 — open weights on Hugging Face
|-
! class="ios-lab" | Sizes
| class="ios-val" | 270M · 450M · 1.1B · 3B parameters (base and instruct)
|-
! class="ios-lab" | Type
| class="ios-val" | Open-source Efficient Language Models (research release)
|}

{{Apple Navbox}}

OpenELM — "Open-source Efficient Language Models" — is Apple's most transparent AI release: a family of small language models published with weights, code, training recipes, and logs, via the company's CoreNet library. Its layer-wise scaling design spends the parameter budget unevenly across the network for better accuracy per parameter at small sizes.

= Position =

OpenELM is a research artifact, not a product — none of its models ship in [[Apple Intelligence]], whose production [[Apple Foundation Models]] are trained separately. Its significance was the gesture: a famously closed company publishing reproducible models sized for on-device use, tuned to run through [[MLX]] on [[Apple silicon]], and signaling that Apple's AI researchers participate in the open ecosystem their products quietly draw on.