Model
Model explorer

Apple OpenELM (270M–3B)

OPEN
Apple · Apple OpenELM family · released Apr 24, 2024

Fully open, reproducible 270M/450M/1.1B/3B family with code, training logs, and checkpoints; layer-wise scaling beats OLMo at matched budget.

ReasoningCodingVisionFunction callingTool useAgentic
-367.9
Elo · rank #437
Parameters
3B
Active params
3B (dense)
Context
2K tokens
Architecture
Dense transformer with layer-wise parameter scaling (270M/450M/1.1B/3B family)
License
Apple Sample Code License
Languages
API price (in/out)
No hosted API
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
ARC-ChallengeReasoning35.6%#157
best: Llama 3.1 405B · 96.9%
ARC-EasyReasoning59.9%#62
best: Phi-3-medium (14B) · 97.7%
HellaSwagReasoning72.4%#118
best: Claude 3 Opus · 95.4%
MMLUKnowledge26.8%#280
best: OpenAI o3 · 92.9%
PIQAReasoning78.2%#54
best: GPT-4o mini · 93.1%
RACE-HReasoning38.8%#22
best: Claude 3 Opus · 92.9%
TruthfulQAKnowledge35.0%#97
best: Phi-3.5-MoE (16x3.8B, 6.6B active) · 77.5%
WinoGrandeReasoning65.5%#117
best: PaLM 2 · 90.9%
Run it locally
VRAM @ Q4
1.9 GB
VRAM @ FP16
6.4 GB
Fits on (Q4)
RTX 3060 12GBRTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB
Throughput data unavailable.
Quantizations
GGUF
Apple OpenELM family
Elo progression across releases
API price weights · each benchmark row carries its own source badge (see methodology)