whatmodel.sh / Llama 3 8B Instruct

Llama 3 8B Instruct

+ contribute

Llama 3 8B Instruct is an open-weights language model from Meta, released April 2024. Its published size is 8.03B parameters and a 8k-token context window. Licence: llama3.

Catalogue identity notes

Comparison model from the Epoch AI internal-evaluation CSVs (Epoch release date 2024-04-18). Original Llama 3 (April 2024) instruct checkpoint Meta-Llama-3-8B-Instruct; distinct from Llama 3.1 8B. 8,030,261,248 parameters (safetensors metadata). ollama llama3:8b = 8b-instruct-q4_0 (same digest).

42.0±0.4#499 of 719 LLMsfrom 45 benchmark results across 25 sources

$ how it compares

The LLMs ranked either side of Llama 3 8B Instruct, out of 828 with a score, then more from Meta.

#	model	score	where	RAM	released
588	Yi-1.5 9B (base) 01.AI	42±2	local	~6.8GB	2024-05
589	Claude Instant 1.2 Anthropic	42±1	comparison only		2023-08
590	Llama 3 8B Instruct Meta	42±0	local	~6.6GB	2024-04
591	WizardLM 70B V1.0 Microsoft	42±1	local	~32GB	2023-08
592	WizardCoder Python 13B V1.0 WizardLM	42±2	local	~9.8GB	2023-08
	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▁▁▃▆█▆▅▅▅▄▃▁ #292 of 300 * · 0.7% of its score · 2 tests

Beats QwQ-32B-Preview, DeepSeek-Coder-V2 Lite Instruct (16B-A2.4B), Phi-4-mini 3.8B 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	33.1% ✎	▁▁▁▃▁▃▅█▅▄▁▁ beats 9/114	medium · avg 52%	▇ 0.4%	blunt · at its level
BFCL (version not stated)	58.9% ✎	█·········██ beats 1/5	easy · avg 78%	▇ 0.4%	blunt · at its level

coding▁▁▁▃▆██▇▆▄▃▁ #342 of 458 · 43% of its score · 9 tests

Beats LFM2.5 2.6B, LFM2.5 8B-A1B, Qwen3-VL 2B Thinking here, though they rank higher overall. Trails Phi-3-mini-128k-instruct (Apr 2024), DeepSeek LLM 67B Chat, OpenChat 3.5 here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
HumanEval	61.6% ✎	▁·▁▁▃▄▆▅█▃▁▁ beats 47/99	easy · avg 79%	▇▇▇▇▇▇ 15%	sharp · at its level
HumanEval+	56.7% ✎	▁·▁▁▁▃▄▄▄█▃▁ beats 48/99	easy · avg 74%	▇▇▇▇▇ 11%	sharp · at its level
CRUXEval (ZeroEval, 800 problems)	37.8% ✎	▁▂▆▃▆█▆▃▂·▁▁ beats 3/47	medium · avg 61%	▇▇ 4.4%	sharp · at its level
HumanEvalSynthesize (HumanEvalPack, 6-language average)	36.9% ✎	▁·▂▁▂▆▅█▄▄▁▂ beats 13/35	medium · avg 64%	▇ 3.0%	at its level
HumanEvalExplain (HumanEvalPack, 6-language average)	29.1% ✎	▄▇█▃▄▇▄▁▇▃·▁ beats 25/35	medium · avg 49%	▇ 2.0%	at its level
LiveCodeBench official (Jan 1–Nov 1, 2024)	13.3% ✎	▁▁·▃▂▄▆█▅▂·▂ beats 2/31	hard · avg 26%	▇ 2.0%	sharp
MBPP+	54.8% ✎	▁···▁▁▂▅▇█▆▂ beats 27/75	easy · avg 68%	▇ 2.0%	at its level
HumanEvalFix (HumanEvalPack, 6-language average)	29.3% ✎	▁·▁▁▄██▄█▁▅▅ beats 24/35	medium · avg 56%	▇ 2.0%	sharp · at its level

1 more test

test	result	vs the others in its run	difficulty	counts	why
WebApp1K-React pass@1 (onekq-ai)	6.8% ✎	▃▁▁█▁▄▅▃▄▇▁▄ beats 0/31	medium · avg 44%	▇ 0.5%	sharp · noisy

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #294 of 393 · 11% of its score · 4 tests

Beats Ministral 3 14B Reasoning 2512, Gemma 4 E4B, OpenThinker3 7B here, though they rank higher overall. Trails Flan-T5 XXL, LFM2 700M, EXAONE 3.5 2.4B Instruct here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
AlpacaEval 2 length-controlled	22.9% ✎	▁▂▂▃▆▆█▆▄▄▁▁ beats 60/122	medium · avg 39%	▇▇ 4.8%	typical
Arena-Hard (v0.1, GPT-4-1106 judge)	20.6% ✎	▂▁▁▅█▅▄▃▃▅▃▁ beats 14/60	medium · avg 52%	▇ 3.1%	sharp
Polyrating rating (bias-aware re-fit of Chatbot Arena votes, mid-2024)	1150.0 ✎	▂▁▂▁▃▇█▆▇▄▂▂ beats 66/95	avg 1201	▇ 1.6%	blunt
LMArena (text)	1223.6 ✎	▁▁▁▂▃▃▃▆▆▇█▃ beats 79/379	avg 1268	▇ 1.6%	blunt

math▁▂▃▅▆▅▄█▄▁▁▁ #364 of 398 · 9.0% of its score · 6 tests

Beats Yi 34B Chat here, though they rank higher overall. Trails Gemma 3 1B, OLMo 2 1B Instruct (0425), Stable Code 3B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
GSM8K (ZeroEval, zero-shot)	78.5% ✎	▁▄·▁▄█▅▅▅███ beats 8/43	very easy · avg 89%	▇ 3.7%	sharp · at its level
GSM8K	72.9% ✎	▂▂▂▅▅▆▅▇█▇▄▃ beats 75/146	very easy · avg 89%	▇ 1.7%	sharp
MATH level 5	6.1% ✎	▂▃▅▆▄█▆▃▅▂▃▄ beats 3/92	hard · avg 32%	▇ 1.3%	sharp
MATH Level 5 (ZeroEval, 721 problems)	7.9% ✎	▁▃▁▆█▂▇▇▄··▂ beats 1/38	hard · avg 34%	▇ 1.2%	sharp
MATH (subset unspecified)	27.9% ✎	▂▄▄▃█▇▅▆▄▃▄▁ beats 12/80	medium · avg 60%	▇ 0.9%	at its level
OTIS mock AIME 2024–2025	1.9% ✎	▃▃▄▃▃▃▇▇█▃▂▅ beats 15/193	very hard · avg 8.5%	▇ 0.1%	sharp · too hard for it · noisy

other tests24 not in a task · 36% of its score

test	result	vs the others in its run	difficulty	counts	why
MMLU-Redux (ZeroEval, 2,778 questions)	61.7% ✎	▁·▁▁▄▄▄▅█▄▂▁ beats 4/53	easy · avg 77%	▇▇▇ 8.5%	at its level
LiveBench 2024-07-26	27.6% ✎	▁▁▁▂▂▃▂█▆▃▅▃ beats 21/61	medium · avg 41%	▇▇ 5.5%	at its level
MMLU-Pro	40.1% ✎	▃▂▅█▃·█▃▅··▂ beats 6/23	medium · avg 57%	▇ 2.6%	at its level
MedQA (USMLE, 4 options)	60.7% ✎	▁▁··▄▁▁··▄▃█ beats 17/17	easy · avg 76%	▇ 2.1%	sharp · at its level
BigCodeBench (Complete)	36.8% ✎	▂▁▁▄▂▅▇█▆▂▂▂ beats 18/58	medium · avg 50%	▇ 1.9%	at its level
WildBench v2 WB-Score (task macro)	29.2028 WB-Score ✎	▁·▁▁·▃█▄▇▆▅▆ beats 11/52	avg 45.199 WB-Score	▇ 1.8%	sharp
BigCodeBench (Instruct)	31.7% ✎	▃▂·▆▆██▆▅▃▃▃ beats 7/33	medium · avg 41%	▇ 1.8%	at its level
BIG-Bench Hard	54.4% ✎	▁▁▁▁▂▄▅▅█▃▄▁ beats 57/129	easy · avg 75%	▇ 1.4%	sharp · at its level · small run

16 more tests

test	result	vs the others in its run	difficulty	counts	why
EvalPerf correctness (118 tasks, 20240328)	43.7% ✎	▃·▂▂▂▆▅▆▆▆▃█ beats 0/29	easy · avg 76%	▇ 1.3%	at its level
CPTUB average (SpeakLeash, 0-5 judge scale)	60.0% ✎	▁▂▂▂▃█▃▆▅▄▂▃ beats 30/95	easy · avg 66%	▇ 1.2%	at its level
MMLU	68.4% ✎	▁▁▁▁▁▂▃▅█▃▆▃ beats 102/183	easy · avg 78%	▇ 1.2%	at its level
MedMCQA	56.9% ✎	▂▂··▄▄▂▂·▄██ beats 15/17	medium · avg 68%	▇ 1.1%	at its level
ZebraLogic (ZeroEval, 1,000 puzzles)	11.9% ✎	▁▁▃▆██▁▁·▁▁▁ beats 14/57	hard · avg 18%	▇ 1.0%	sharp · too hard for it
Polish EQ-Bench v2 (SpeakLeash)	46.3% ✎	▁··▁▁▂▂▆▅▆█▆ beats 20/70	medium · avg 64%	▇ 1.0%	at its level
GPQA Diamond	26.1% ✎	▁▃▁▁▂▄▆▄▅▇█▅ beats 6/208	hard · avg 40%	▇ 0.9%	sharp · too hard for it
Open Persian LLM Leaderboard v2 average (21 tasks)	50.4% ✎	▂▄▂▃█▄▃▆▃▅▂▂ beats 32/81	medium · avg 58%	▇ 0.7%	blunt · at its level
GPQA (subset unspecified)	26.1% ✎	▂▃▂▂▄▃▆▃▅█▄▃ beats 6/104	hard · avg 40%	▇ 0.6%	sharp
AraGen-v3 3C3H score (Inception, Claude Sonnet 4 judge)	9.2% ✎	▁▁·▁▂▃▄█▄▅▄▄ beats 6/66	hard · avg 31%	▇ 0.5%	typical
Open PT-LLM Leaderboard average (9 Portuguese tasks)	50.1% ✎	▃▁▁▁▁▁▁▃▂█▅▁ beats 67/276	medium · avg 73%	▇ 0.4%	at its level · noisy
CMMLU	53.3% ✎	▅·▅····█▃▅█▅ beats 1/14	easy · avg 75%	▇ 0.3%	at its level · noisy
EuroEval Life in the UK (English knowledge MC), test split, 5-shot, accuracy	83.5% ✎	▁▁▁▁▁▁▁▂▄█▄▃ beats 87/146	easy · avg 83%	▇ 0.2%	at its level · noisy
Vectara hallucination leaderboard, 2025 dataset (HHEM-2.3, factual consistency rate)	92.6% ✎	▁▁▂▃▄▄█▆▃▂▂▁ beats 40/140	very easy · avg 94%	▇ 0.1%	blunt · noisy
PubMedQA	74.6% ✎	▂·▂▆·▄▂▄▄▂█▂ beats 13/17	medium · avg 77%	▇ 0.1%	blunt · at its level
Chess Puzzles (Epoch)	0.0% ✎	█▁▁▁▁▂▁▄▂▂▁▁ beats 7/135	very hard · avg 5.9%	▇ 0.0%	sharp · too hard for it · noisy

$ where to run it

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 8B Instruct · needs ~4.9 GB · from huggingface.co/lmstudio-community
chat in the terminal
llama cli -hf lmstudio-community/Meta-Llama-3-8B-Instruct-GGUF:Q4_K_M
Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf lmstudio-community/Meta-Llama-3-8B-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 8B Instruct · needs ~4.9 GB · from huggingface.co/lmstudio-community
open in LM Studio ↗
Or download it in the terminal
lms get https://huggingface.co/lmstudio-community/Meta-Llama-3-8B-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 8B Instruct · needs ~5.32 GB · from ollama.com
chat in the terminal
ollama run llama3:8b

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 8B Instruct · needs ~5.3 GB, Mac only · from huggingface.co/mlx-community
chat in the terminal
mlx_lm.chat --model mlx-community/Meta-Llama-3-8B-Instruct-4bit
Or start an OpenAI-compatible server at http://localhost:8080/v1
mlx_lm.server --model mlx-community/Meta-Llama-3-8B-Instruct-4bit

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

$ how the score was worked out · EvalPlus / EvalPerf published… 30%, WildEval ZeroEval published r… 19%, Granite Code Models: A Family… 7% and 22 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
EvalPlus / EvalPerf published evaluation runs	4	▇▇▇▇▇▇▇▇▇ 30%	Apache-2.0
WildEval ZeroEval published results	5	▇▇▇▇▇▇ 19%	Apache-2.0
Granite Code Models: A Family of Open Foundation Models for Code Intelligence — authors' own runs	3	▇▇ 7%	CC BY 4.0
InternLM GitHub repository model cards (Shanghai AI Laboratory)	6	▇▇ 6%	Apache-2.0
LiveBench model judgments (HF livebench/model_judgment, 2024-10-22 revision) with the LiveBench question datasets	1	▇▇ 6%	Apache-2.0 (stated in the LiveBench datasheet)
AlpacaEval 1.0 and 2.0 leaderboards (tatsu-lab/alpaca_eval repository)	1	▇ 5%	Apache-2.0
BigCodeBench paper (Zhuo et al., ICLR 2025), arXiv 2406.15877v4	2	▇ 4%	CC BY 4.0
Open Medical-LLM Leaderboard results (HF dataset openlifescienceai/results)	3	▇ 3%	Apache-2.0
Arena-Hard-Auto v0.1 leaderboard (repository README, pre-v2)	1	▇ 3%	Apache-2.0
MMLU-Pro author-released evaluation runs	1	▇ 3%	Apache-2.0
Epoch AI — internal evaluations	4	▇ 2%	CC BY 4.0
LiveCodeBench official website results (2024-12-26)	1	▇ 2%	CC-BY-SA-4.0
WildBench v2 leaderboard data (Ai2)	1	▇ 2%	Apache-2.0
Polyrating: A Cost-Effective and Bias-Aware Rating System for LLM Evaluation (arXiv:2409.00696v3)	1	▇ 2%	CC-BY-4.0
LMArena leaderboard dataset	1	▇ 2%	CC BY 4.0
SpeakLeash CPTUB (Complex Polish Text Understanding Benchmark)	1	▇ 1%	MIT
SpeakLeash Polish EQ-Bench leaderboard	1	▇ 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)	1	▇ 1%	Apache-2.0
WebApp1K models leaderboard (onekq-ai)	1	▇ 0%	MIT
Open Portuguese LLM Leaderboard (archived)	1	▇ 0%	Apache-2.0
Berkeley Function Calling Leaderboard result archive	1	▇ 0%	Apache-2.0
ChatGLM: A Family of Large Language Models from GLM-130B to GLM-4 All Tools (Team GLM, arXiv 2406.12793v2)	1	▇ 0%	CC BY 4.0
EuroEval leaderboards (results collection), English generative leaderboard	1	▇ 0%	MIT
Vectara hallucination leaderboard (vectara/hallucination-leaderboard repository)	1	▇ 0%	Apache-2.0
prior · pull toward the middle		▇ 0%
weighted average		= 42.0