Model
Model explorer

SmolLM2 1.7B

OPEN
Hugging Face · SmolLM family · released Nov 1, 2024

Fully-open small-model family (135M/360M/1.7B) for on-device use, trained on 11T curated tokens (FineWeb-Edu, DCLM, Stack, custom math/code).

ReasoningCodingVisionFunction callingTool useAgentic
-93.8
Elo · rank #411
Parameters
1.71B
Active params
1.71B (dense)
Context
8K tokens
Architecture
Llama-style decoder-only Transformer (SwiGLU, RoPE, tied embeddings), 24 layers, 2048 hidden dim
License
Apache-2.0
Languages
1+
API price (in/out)
No hosted API
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
ARC-ChallengeReasoning60.5%#88
best: Llama 3.1 405B · 96.9%
GSM8KMath31.0%#153
best: Llama 3.1 405B · 96.8%
HellaSwagReasoning68.7%#130
best: Claude 3 Opus · 95.4%
MMLU-ProKnowledge19.4%#171
best: Claude Fable 5 · 91.5%
OpenBookQAReasoning42.2%#41
best: Claude 1 · 90.8%
PIQAReasoning77.6%#57
best: GPT-4o mini · 93.1%
TriviaQAKnowledge36.7%#46
best: Sarvam-1 (2B) · 90.6%
WinoGrandeReasoning59.4%#132
best: PaLM 2 · 90.9%
Run it locally
VRAM @ Q4
1.5 GB
VRAM @ FP16
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 Q4 · GGUF Q8 · MLX
SmolLM family
Elo progression across releases
SmolLM2 1.7BNov 2024-93.8SmolLM3 3B (No Thinking)Jul 2025703.6SmolLM3 3B (Think)Jul 2025881.6
Resources
HFHuggingFaceTB/SmolLM2-1.7BGHhuggingface/smollmDOCSarxiv.org/abs/2502.02737
API price weights · each benchmark row carries its own source badge (see methodology)