whatmodel.sh / Llama 3.2 1B Instruct

Llama 3.2 1B Instruct

+ contribute

Llama 3.2 1B Instruct is an open-weights language model from Meta, released September 2024. Its published size is 1.24B parameters and a 131k-token context window. Licence: llama3.2.

29.8±0.6#642 of 719 LLMsfrom 65 benchmark results across 19 sources

$ how it compares

The LLMs ranked either side of Llama 3.2 1B Instruct, out of 828 with a score, then more from Meta.

#	model	score	where	RAM	released
737	LFM2 350M Liquid AI	30±2	local	~1.7GB	2025-07
738	SmolLM2 1.7B Instruct Hugging Face	30±1	local	~2.6GB	2024-10
739	Llama 3.2 1B Instruct Meta	30±1	api $0.020/M +local	~2.2GB	2024-09
740	Falcon 7B (base) TII	30±1	local	~5.9GB	2023-04
741	Llama 2 7B Chat Meta	30±1	local	~5.8GB	2023-07
	more from Meta
24	Muse Spark 1.3 Meta	77±1	api $2.00/M		2026-09
40	Muse Spark 1.2 Meta	75±1	api $2.00/M		2026-08
43	Muse Spark 1.1 Meta	75±1	api $2.00/M		2026-07

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▃▁▃▆█▆▅▅▅▄▃▁ #296 of 300 · 2.3% of its score · 3 tests

Beats QwQ-32B-Preview, DeepSeek-Coder-V2 Lite Instruct (16B-A2.4B), Gemma 3 1B here, though they rank higher overall. Trails Granite 4.0 350M, Qwen2.5 0.5B Instruct, Granite 4.0 H 350M here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
BFCL v3 +1	21.9% ✎	▁▃▁▁▁▃▅█▅▄▁▁ beats 5/114	medium · avg 52%	▇ 1.1%	blunt · at its level · several runs
BFCL V4 (December 2025 archive)	10.8% ✎	▁▃▁▂██▅▅▆▂▂▂ beats 1/77	hard · avg 32%	▇ 0.8%	typical
MASSIVE-Agents AST (10k, zero-shot)	0.2% ✎	█····███·███ beats 0/6	hard · avg 16%	▇ 0.5%	sharp · too hard for it

coding▁▁▁▃▆██▇▆▄▃▁ #407 of 458 · 8.7% of its score · 6 tests

Beats DeepSeek R1 Distill Qwen 7B, Qwen3.5 2B, DeepSeek R1 Distill Llama 8B here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
HumanEval +1	36.0% ✎	▄·▄█·▄·▄···█ beats 1/7	easy · avg 81%	▇▇ 6.0%	sharp · at its level · several runs
MBPP	37.0% ✎	▄···▄·▄▄█··█ beats 3/7	medium · avg 64%	▇ 1.8%	at its level
CodeReviewQA: automated code refinement, exact match rate	5.7% ✎	▁▁▁▃▁▁▂▂▃█▅▄ beats 4/47	hard · avg 33%	▇ 0.4%	typical
CodeReviewQA: solution identification (hard), invariant accuracy	0.8% ✎	▃··▁▂▂▃▂█▃▄▂ beats 4/47	medium · avg 54%	▇ 0.3%	sharp · noisy
CodeReviewQA: change localisation (hard), invariant accuracy	0.0% ✎	█▂▁▁▂▂▂▃▅▅▅▇ beats 5/47	medium · avg 36%	▇ 0.2%	sharp · too hard for it · noisy
CodeReviewQA: change type recognition, invariant accuracy	76.1% ✎	▁▁▁······▁▃█ beats 34/47	easy · avg 67%	▇ 0.0%	blunt · at its level · noisy

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #368 of 393 · 12% of its score · 6 tests

Beats Code Llama 70B Instruct, Phi-2 (2.7B), Qwen2.5 1.5B Instruct here, though they rank higher overall. Trails Gemma 3 270M IT, SmolLM2 360M Instruct, Granite 4.0 350M here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
LMArena (text)	1111.1 ✎	▁▁▁▃▃▃▃▆▆▇█▃ beats 18/379	avg 1268	▇ 3.6%	blunt
IFEval (aggregation unspecified) +2	41.7% ✎	▄····█·▄··██ beats 1/7	easy · avg 82%	▇ 3.5%	at its level · several runs
IFEval mean of four measures	52.4% ✎	█·█··█·██·██ beats 0/6	very easy · avg 87%	▇ 1.9%	at its level
IFEval prompt-level loose	54.0% ✎	▁▃▁▃▁▄▁▁█▃▃▁ beats 2/22	easy · avg 80%	▇ 1.0%	at its level
Arabic IFEval prompt-level (Inception)	11.1% ✎	▁·▁▁▂▃▄▄▆▄█▄ beats 5/38	medium · avg 49%	▇ 0.9%	sharp · too hard for it
AlpacaEval 2 length-controlled	10.1% ✎	▃▃·▁▁█▇▅·▁·▁ beats 3/22	medium · avg 42%	▇ 0.7%	too hard for it

math▁▃▃▅▆▅▄█▄▁▁▁ #377 of 398 · 25% of its score · 6 tests

Beats Command R (03-2024), Molmo 7B-D (0924), Mistral 7B Instruct v0.3 here, though they rank higher overall. Trails Gemma 2B (base) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
GSM8K +4	45.4% ✎	▂▃··▃▂▅█▆▃▂▃ beats 2/22	very easy · avg 88%	▇▇▇▇▇▇ 15%	sharp · at its level · several runs
MATH (subset unspecified) +2	21.6% ✎	▃▃█▂▂▆·····▁ beats 2/22	medium · avg 53%	▇▇ 4.9%	several runs
MGSM	29.1% ✎	█···█▄··▄··▄ beats 0/6	easy · avg 77%	▇ 3.3%	at its level
MATH-500	23.4% ✎	▄··█▄·▄···▄▄ beats 0/6	easy · avg 80%	▇ 1.2%	at its level
AIME 2025	0.3% ✎	▄█··▄···▄··█ beats 0/6	hard · avg 33%	▇ 0.4%	sharp · too hard for it · noisy
OTIS mock AIME 2024–2025	0.6% ✎	▃▃▄▃▃▃▇▇█▃▂▅ beats 5/193	very hard · avg 8.5%	▇ 0.0%	sharp · too hard for it · noisy

other tests26 not in a task · 52% of its score

test	result	vs the others in its run	difficulty	counts	why
MMLU +2	46.7% ✎	▁▂·▃·▁▁█·▃▃▅ beats 3/22	easy · avg 77%	▇▇▇▇ 10%	several runs
BIG-Bench Hard +2	40.2% ✎	▁▄·▁▁·▅█·▄▄▁ beats 3/22	medium · avg 67%	▇▇▇ 8.7%	sharp · several runs
HellaSwag +1	59.7% ✎	▄·▄··█▄·▄·▄▄ beats 0/7	easy · avg 87%	▇ 3.7%	at its level · several runs
WinoGrande	61.0% ✎	█···▄▄▄▄···█ beats 0/7	easy · avg 84%	▇ 3.5%	at its level
MMLU-Pro +1	19.5% ✎	▄·····█··█▄█ beats 1/7	medium · avg 47%	▇ 2.7%	several runs
BoolQ	66.7% ✎	▄▄····▄▄·█·█ beats 0/7	easy · avg 91%	▇ 2.5%	at its level
CPTUB average (SpeakLeash, 0-5 judge scale)	38.4% ✎	▁▃▂▂▃█▃▆▅▄▂▃ beats 6/95	easy · avg 66%	▇ 2.5%	at its level
EQ-Bench 3 (emotional intelligence, judge Elo)	856.9 ✎	▃·▂▁▄▅▅▃█▄▁▁ beats 0/62	avg 1044	▇ 2.2%	typical

18 more tests

test	result	vs the others in its run	difficulty	counts	why
GPQA (subset unspecified) +1	25.7% ✎	▄·▄··▄··▄█·█ beats 1/7	hard · avg 39%	▇ 2.2%	sharp · too hard for it · several runs
IFBench (aggregation unspecified) +1	15.9% ✎	▄·█·█······█ beats 0/6	medium · avg 40%	▇ 2.0%	several runs
Polish EQ-Bench v2 (SpeakLeash)	20.6% ✎	▁··▁▃▂▂▆▅▆█▆ beats 3/70	medium · avg 64%	▇ 1.5%	typical
Open Persian LLM Leaderboard v2 average (21 tasks)	35.0% ✎	▂▄▂▃█▄▃▆▃▅▂▂ beats 6/81	medium · avg 58%	▇ 1.4%	blunt · at its level
MMMLU	38.1% ✎	▄···▄·▄▄··▄█ beats 3/6	medium · avg 59%	▇ 1.3%	typical
ARC Challenge	41.4% ✎	▄····█··▄▄▄█ beats 2/7	easy · avg 78%	▇ 1.2%	sharp · noisy
BRIDGE clinical text, zero-shot average (87 tasks)	12.7% ✎	▁▃▁▁▁▂▂▃▄▆█▃ beats 1/97	hard · avg 32%	▇ 1.2%	blunt
OpenBookQA	34.6% ✎	▄▄··▄··▄·█▄▄ beats 0/7	medium · avg 60%	▇ 1.1%	typical
TruthfulQA MC2	41.5% ✎	▃▃▂·▃▃·▅█▂▅▂ beats 0/21	medium · avg 63%	▇ 0.8%	blunt · at its level
Open PT-LLM Leaderboard average (9 Portuguese tasks)	35.9% ✎	▃▁▁▁▃▁▁▁▂█▅▁ beats 44/276	medium · avg 73%	▇ 0.6%	noisy
PIQA	72.3% ✎	▁▁▁▃▃▄▇█▄▃▁▁ beats 21/83	easy · avg 86%	▇ 0.5%	blunt · at its level · small run
EuroEval Life in the UK (English knowledge MC), test split, 5-shot, accuracy	32.4% ✎	▁▁▁▁▃▁▁▂▄█▄▃ beats 7/146	easy · avg 83%	▇ 0.5%	at its level · noisy
PopQA	13.8% ✎	▁▁▃▃▃▃▁█▁▃·▄ beats 2/21	hard · avg 32%	▇ 0.5%	blunt
Vectara hallucination leaderboard, 2025 dataset (HHEM-2.3, factual consistency rate)	79.3% ✎	▁▃▂▃▄▄█▆▃▂▂▁ beats 4/140	very easy · avg 94%	▇ 0.5%	blunt · noisy
AraGen-v3 3C3H score (Inception, Claude Sonnet 4 judge)	10.5% ✎	▁▁·▁▂▃▄█▄▅▄▄ beats 7/66	hard · avg 31%	▇ 0.5%	too hard for it
GPQA Diamond +1	23.9% ✎	▃▁▁▁▂▄▆▄▅▇█▅ beats 5/208	hard · avg 40%	▇ 0.4%	sharp · too hard for it
Social IQA	35.5% ✎	▃·····▃··█▃▅ beats 0/7	medium · avg 68%	▇ 0.4%	blunt · at its level
Chess Puzzles (Epoch)	0.0% ✎	█▁▁▁▁▂▁▄▂▂▁▁ beats 37/135	very hard · avg 5.9%	▇ 0.0%	sharp · too hard for it · noisy

$ where to run it

Through an API

api · Novita AI · $0.020/M · $0.02 in / $0.02 out per 1M tokens · 131k context · novita.ai ↗ · price sourceUSD per 1M text tokens, standard uncached input and output. Model endpoint meta-llama/llama-3.2-1b-instruct. Product: Serverless chat completions. API model: meta-llama/llama-3.2-1b-instruct. Settings: service tier=standard, batch=no, cache=uncached input. Additional/excluded charges: Tools, cache storage, priority service and other modalities excluded.. Currency: USD.

llama.cpp · recommended

version

Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
macOS / Linux
curl -LsSf https://llama.app/install.sh | sh
Windows
irm https://llama.app/install.ps1 | iex
other ways to install ↗
Run Llama 3.2 1B Instruct · needs ~0.8 GB · from huggingface.co/lmstudio-community
chat in the terminal
llama cli -hf lmstudio-community/Llama-3.2-1B-Instruct-GGUF:Q4_K_M
Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf lmstudio-community/Llama-3.2-1B-Instruct-GGUF:Q4_K_M

LM Studio

version

Install LM Studio · Open the app once before using the lms command.
download it (macOS / Windows / Linux) ↗
Run Llama 3.2 1B Instruct · needs ~0.8 GB · from huggingface.co/lmstudio-community
open in LM Studio ↗
Or download it in the terminal
lms get https://huggingface.co/lmstudio-community/Llama-3.2-1B-Instruct-GGUF@Q4_K_M -y
Then run lms ls to see its name, and lms chat followed by that name

Ollama

version

Install Ollama
download it (macOS / Windows) ↗
Linux
curl -fsSL https://ollama.com/install.sh | sh
Run Llama 3.2 1B Instruct · needs ~1.99 GB · from ollama.com
chat in the terminal
ollama run llama3.2:1b

MLX (Mac)

version

Install MLX (Mac) · Apple silicon Macs only, text only.
with uv
uv tool install mlx-lm
or
pip install mlx-lm
Run Llama 3.2 1B Instruct · needs ~0.7 GB, Mac only · from huggingface.co/rajib-sarwar
chat in the terminal
mlx_lm.chat --model rajib-sarwar/Llama-3.2-1B-Instruct-4bit-MLX
Or start an OpenAI-compatible server at http://localhost:8080/v1
mlx_lm.server --model rajib-sarwar/Llama-3.2-1B-Instruct-4bit-MLX

Weights and documentation: huggingface.co/meta-llama ↗

$ how the score was worked out · IBM Granite 3.0 Language Mode… 34%, 2 OLMo 2 Furious (Team OLMo,… 15%, LiquidAI/LFM2-1.2B model card 12% and 16 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

source	results	share of the score	licence
IBM Granite 3.0 Language Models technical report (GitHub)	15	▇▇▇▇▇▇▇▇▇▇ 34%	Apache-2.0
2 OLMo 2 Furious (Team OLMo, Ai2) — authors' own evaluations of OLMo 2 Instruct and peer instruct models	8	▇▇▇▇ 15%	CC BY 4.0
LiquidAI/LFM2-1.2B model card	7	▇▇▇▇ 12%	LFM Open License v1.0
SmolLM2 paper (Hugging Face) — authors' own instruct-model evaluations	7	▇▇▇ 11%	CC BY 4.0
LiquidAI/LFM2.5-1.2B-Thinking model card	7	▇▇▇ 9%	LFM Open License v1.0
LMArena leaderboard dataset	1	▇ 4%	CC BY 4.0
SpeakLeash CPTUB (Complex Polish Text Understanding Benchmark)	1	▇ 2%	MIT
EQ-Bench 3 canonical leaderboard results	1	▇ 2%	MIT
SpeakLeash Polish EQ-Bench leaderboard	1	▇ 2%	Apache-2.0
Berkeley Function Calling Leaderboard result archive	2	▇ 1%	Apache-2.0
Open Persian LLM Leaderboard v2 (Part DP AI and AUT NLP Lab)	1	▇ 1%	Apache-2.0
Inception Arabic/Hindi leaderboards (AraGen-v3, HindiGen-v1, Arabic IFEval)	2	▇ 1%	Apache-2.0
BRIDGE Medical Leaderboard (Harvard Medical School / Brigham and Women's YLab)	1	▇ 1%	MIT
CodeReviewQA: The Code Review Comprehension Assessment for Large Language Models (arXiv:2503.16167v2)	4	▇ 1%	CC-BY-4.0
Open Portuguese LLM Leaderboard (archived)	1	▇ 1%	Apache-2.0
EuroEval leaderboards (results collection), English generative leaderboard	1	▇ 1%	MIT
MASSIVE-Agents — Kulkarni et al. (2025)	1	▇ 1%	CC BY 4.0
Vectara hallucination leaderboard (vectara/hallucination-leaderboard repository)	1	▇ 0%	Apache-2.0
Epoch AI — internal evaluations	3	▇ 0%	CC BY 4.0
prior · pull toward the middle		▇ 0%
weighted average		= 29.8