Claude Fable 5$22.000/MClaude Opus 4.8$11.000/MClaude Opus 4.7$11.000/MClaude Opus 4.6$11.000/MClaude Opus 4.5$33.000/MClaude Sonnet 3.7$6.600/MClaude Opus 3$33.000/MClaude 2.1$12.800/MClaude 2$12.800/MGPT-5.5$12.500/MGPT-5.2$5.425/MGPT-5.2-Codex$5.425/MGPT-5$3.875/MGPT-4.5$97.500/MGPT-4 Turbo Preview$16.000/MGPT-4$39.000/MGPT-4-32k$78.000/Mo3$19.000/Mo3-mini$2.090/Mo4-mini$2.090/Mo1$28.500/Mo1-mini$5.700/Mo1-preview$28.500/MGemini 3.5 Pro$5.000/MGemini 3.1 Pro$5.000/MGemini 3 Pro$5.000/MGemini 2.5 Pro$3.875/MGemini 1.5 Pro$2.375/MGemini 1.0 Ultra$12.000/MGemini 1.0 Pro$0.800/MClaude Fable 5$22.000/MClaude Opus 4.8$11.000/MClaude Opus 4.7$11.000/MClaude Opus 4.6$11.000/MClaude Opus 4.5$33.000/MClaude Sonnet 3.7$6.600/MClaude Opus 3$33.000/MClaude 2.1$12.800/MClaude 2$12.800/MGPT-5.5$12.500/MGPT-5.2$5.425/MGPT-5.2-Codex$5.425/MGPT-5$3.875/MGPT-4.5$97.500/MGPT-4 Turbo Preview$16.000/MGPT-4$39.000/MGPT-4-32k$78.000/Mo3$19.000/Mo3-mini$2.090/Mo4-mini$2.090/Mo1$28.500/Mo1-mini$5.700/Mo1-preview$28.500/MGemini 3.5 Pro$5.000/MGemini 3.1 Pro$5.000/MGemini 3 Pro$5.000/MGemini 2.5 Pro$3.875/MGemini 1.5 Pro$2.375/MGemini 1.0 Ultra$12.000/MGemini 1.0 Pro$0.800/M
BETA
Subscribe
← Back to Index·All Meta models
Meta
MetaEfficient

Llama 3.2 3B

Micro

Edge-targeted Llama 3.2 — designed for on-device assistants on phones and laptops.

Llama 3.2 3B is a efficient AI model from Meta. It costs $0.060 per million input tokens and $0.060 per million output tokens (blended $0.060/M), with a 128K-token context window.

Released Sep 2024Modalities textOfficial model page ↗API docs ↗Compare with another model →Estimate monthly cost →
INPUT
$0.060/M
per million input tokens
OUTPUT
$0.060/M
per million output tokens
BLENDED 70/30
$0.060/M
default reference rate · how it's calculated →
CONTEXT
128K
128,000 tokens
What it's good at
  • Runs on-device
  • Open weights
  • 128K context
Typical use cases
  • Mobile/edge assistants
  • On-device summarisation
Benchmarks
vs. best public score
MMLU63%
Multitask academic knowledge across 57 subjects.
GPQA Diamond27%
Graduate-level science questions, "Google-proof".
MATH48%
High-school competition math problems.
HumanEval50%
Python function synthesis from docstrings.
Hand-curated from each provider's published reports and public leaderboards. Methodology varies across sources — treat as directional rather than authoritative.
More from Meta
See all 19 →
Llama 4 Maverick
Frontier · 128K ctx
in $0.200/Mout $0.600/M
Llama 4 Scout
Efficient · 512K ctx
in $0.100/Mout $0.350/M
Llama 3.3 70B
Open · 128K ctx
in $0.230/Mout $0.400/M
Llama 3.1 405B
Large · 128K ctx
in $0.800/Mout $0.800/M
Llama 3.1 70B
Balanced · 128K ctx
in $0.350/Mout $0.400/M
Llama 3.1 8B
Compact · 128K ctx
in $0.100/Mout $0.100/M
Frequently asked questions

How much does Llama 3.2 3B cost?

Llama 3.2 3B costs $0.060 per million input tokens and $0.060 per million output tokens, for a blended reference rate of $0.060 per million tokens.

What is Llama 3.2 3B's context window?

Llama 3.2 3B supports up to 128K tokens of context (128,000 tokens).

What is Llama 3.2 3B best for?

Llama 3.2 3B is well suited to Runs on-device, Open weights and 128K context.

Who makes Llama 3.2 3B?

Llama 3.2 3B is developed and served by Meta. It was released in Sep 2024.