whatmodel.sh / Llama 3 70B Instruct

Llama 3 70B Instruct

+ contribute

Llama 3 70B Instruct is an open-weights language model from Meta, released April 2024. Its published size is 70.55B 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-3-70B-Instruct; distinct from Llama 3.1/3.3 70B. 70,553,706,496 parameters (safetensors metadata). ollama llama3:70b = 70b-instruct-q4_0 (same digest).

49.6±0.4#350 of 719 LLMsfrom 34 benchmark results across 23 sources

$ how it compares

The LLMs ranked either side of Llama 3 70B Instruct, out of 828 with a score, then more from Meta.

#	model	score	where	RAM	released
419	Codestral 22B v0.1 Mistral AI	50±1	local	~15GB	2024-05
420	GLM-4 (0116) Z.ai	50±1	comparison only		2024-01
421	Llama 3 70B Instruct Meta	50±0	local	~24GB	2024-04
422	WizardLM-2 8x22B Microsoft	50±1	local	~91GB	2024-04
423	Nova Pro Amazon	50±1	api $1.40/M		2024-12
	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▁▁▃▆█▆▅▅▅▄▃▁ #251 of 300 * · 0.4% of its score · 1 test

Beats DeepSeek R1 0528 Qwen3 8B, Magistral Small 1.0 (2506), Llama 4 Scout (109B-A17B) here, though they rank higher overall. Trails Granite 4.0 H 1B, Granite 4.0 1B, SmolLM3 3B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
BFCL v3	50.8% ✎	▁▁▁▁▁▃▅█▅▄▁▁ beats 37/114	medium · avg 52%	▇ 0.4%	blunt · at its level

coding▁▁▁▃▆██▇▆▄▃▁ #267 of 458 · 42% of its score · 10 tests

Beats Qwen3-VL 32B Instruct, Solar Open 100B (102B-A12B), GLM-4.7-Flash here, though they rank higher overall. Trails DeepSeek LLM 67B Chat, Mistral NeMo Instruct (2407), QwQ-32B-Preview here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
HumanEval	77.4% ✎	▁·▁▁▃▄▆▅█▃▁▁ beats 81/99	easy · avg 79%	▇▇▇▇▇ 12%	sharp
HumanEval+	72.0% ✎	▁·▁▁▁▃▄▄▄█▃▁ beats 80/99	easy · avg 74%	▇▇▇▇ 10%	sharp · at its level
CRUXEval (ZeroEval, 800 problems)	58.9% ✎	▁▂▆▃▆█▆▃▂·▁▁ beats 18/47	medium · avg 61%	▇▇ 5.1%	sharp · at its level
LiveCodeBench official (Jan 1–Nov 1, 2024)	24.3% ✎	▁▁·▁▂▄▆█▅▂·▂ beats 9/31	hard · avg 26%	▇▇ 3.8%	sharp
HumanEvalSynthesize (HumanEvalPack, 6-language average)	65.5% ✎	▁·▂▁▂▆▅█▄▄▁▃ beats 35/35	medium · avg 64%	▇ 3.1%	at its level
HumanEvalExplain (HumanEvalPack, 6-language average)	45.4% ✎	▄▇█▃▄▇▄▁▇▃·▁ beats 34/35	medium · avg 49%	▇ 2.6%	at its level
HumanEvalFix (HumanEvalPack, 6-language average)	52.4% ✎	▁·▁▁▄██▄█▁▅▅ beats 33/35	medium · avg 56%	▇ 2.5%	sharp · at its level
MBPP+	69.0% ✎	▁···▁▁▂▅▇█▆▂ beats 60/75	easy · avg 68%	▇ 2.0%	at its level

2 more tests

test	result	vs the others in its run	difficulty	counts	why
WebApp1K-React pass@1 (onekq-ai)	33.2% ✎	▃▁▁█▁▄▅▃▄▇▁▄ beats 8/31	medium · avg 44%	▇ 0.9%	sharp · at its level · noisy
SciCode main problems (official leaderboard)	0.0% ✎	▃···█··▂·▃·▂ beats 0/17	very hard · avg 1.1%	▇ 0.5%	sharp · too hard for it

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #258 of 393 · 16% of its score · 4 tests

Beats Ministral 3 14B Reasoning 2512, Ministral 3 14B, OpenThinker2 32B here, though they rank higher overall. Trails Flan-T5 XXL, Granite 4.0 H 1B, Granite 4.0 1B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
AlpacaEval 2 length-controlled	34.4% ✎	▁▂▂▃▆▆█▆▄▄▁▁ beats 87/122	medium · avg 39%	▇▇▇ 7.5%	at its level
Arena-Hard (v0.1, GPT-4-1106 judge)	46.6% ✎	▂▁▁▅█▅▄▃▃▅▃▁ beats 35/60	medium · avg 52%	▇▇ 5.0%	sharp · at its level
Polyrating rating (bias-aware re-fit of Chatbot Arena votes, mid-2024)	1187.0 ✎	▂▁▂▁▃▇█▆▇▄▂▂ beats 79/95	avg 1201	▇ 1.8%	blunt
LMArena (text)	1276.5 ✎	▁▁▁▂▃▃▃▆▆▇█▃ beats 109/379	avg 1268	▇ 1.8%	blunt

math▁▂▃▅▆▅▄█▄▁▁▁ #310 of 398 · 9.0% of its score · 4 tests

Beats Athene-70B here, though they rank higher overall. Trails Gemma 7B (base), DeepSeek R1 Distill Qwen 1.5B, Tulu 2 DPO 13B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
MATH level 5	22.6% ✎	▂▂▅▆▄█▆▃▅▂▃▄ beats 18/92	hard · avg 32%	▇▇ 3.9%	sharp · at its level
MATH Level 5 (ZeroEval, 721 problems)	25.1% ✎	▁▁▁▆█▂▇▇▄··▂ beats 13/38	hard · avg 34%	▇ 2.4%	sharp · at its level
GSM8K (ZeroEval, zero-shot)	93.0% ✎	▁▄·▁▄█▅▅▅███ beats 28/43	very easy · avg 89%	▇ 2.2%	sharp
OTIS mock AIME 2024–2025	4.3% ✎	▃▃▄▃▃▃▇▇█▃▂▅ beats 27/193	very hard · avg 8.5%	▇ 0.5%	sharp · too hard for it · noisy

other tests15 not in a task · 32% of its score

test	result	vs the others in its run	difficulty	counts	why
MMLU-Redux (ZeroEval, 2,778 questions)	78.0% ✎	▁·▁▁▄▄▄▅█▄▂▁ beats 25/53	easy · avg 77%	▇▇▇ 8.1%	at its level
LiveBench 2024-07-26	37.7% ✎	▁▁▁▂▂▃▂█▆▃▅▃ beats 41/61	medium · avg 41%	▇▇▇ 7.7%	at its level
ZebraLogic (ZeroEval, 1,000 puzzles)	16.8% ✎	▁▁▃▆██▁▁·▁▁▁ beats 23/57	hard · avg 18%	▇ 2.6%	sharp
BigCodeBench (Instruct)	43.2% ✎	▃▂·▆▆██▆▅▃▃▃ beats 29/33	medium · avg 41%	▇ 2.3%	at its level
BigCodeBench (Complete)	53.9% ✎	▂▁▁▄▂▅▇█▆▂▃▂ beats 55/58	medium · avg 50%	▇ 2.2%	at its level
GPQA Diamond	40.6% ✎	▁▁▁▁▂▄▆▄▅▇█▅ beats 39/208	hard · avg 40%	▇ 2.0%	sharp
WildBench v2 WB-Score (task macro)	47.7708 WB-Score ✎	▁·▁▁·▃█▄▇▆▅▆ beats 38/52	avg 45.199 WB-Score	▇ 2.0%	sharp
CPTUB average (SpeakLeash, 0-5 judge scale)	75.6% ✎	▁▂▂▂▃█▃▆▅▄▂▃ beats 71/95	easy · avg 66%	▇ 1.2%	at its level

7 more tests

test	result	vs the others in its run	difficulty	counts	why
Polish EQ-Bench v2 (SpeakLeash)	71.2% ✎	▁··▁▁▂▂▆▅▆█▆ beats 58/70	medium · avg 64%	▇ 1.0%	at its level
AraGen-v3 3C3H score (Inception, Claude Sonnet 4 judge)	26.1% ✎	▁▁·▁▂▂▄█▄▅▄▄ beats 21/66	hard · avg 31%	▇ 0.8%	at its level
Open Persian LLM Leaderboard v2 average (21 tasks)	65.9% ✎	▂▄▂▃█▄▃▆▃▅▂▂ beats 66/81	medium · avg 58%	▇ 0.8%	blunt · at its level
KINA (Knowledge Index of Noah's Ark) accuracy, 899 ten-option items	26.6% ✎	▁▁▂▁▃▄█▆▄▅▂▃ beats 12/39	very hard · avg 22%	▇ 0.7%	blunt
Open PT-LLM Leaderboard average (9 Portuguese tasks)	78.5% ✎	▃▁▁▁▁▁▁▁▂█▅▁ beats 244/276	medium · avg 73%	▇ 0.4%	at its level · noisy
EuroEval Life in the UK (English knowledge MC), test split, 5-shot, accuracy	94.3% ✎	▁▁▁▁▁▁▁▂▄█▄▃ beats 135/146	easy · avg 83%	▇ 0.2%	noisy
Vectara hallucination leaderboard, 2025 dataset (HHEM-2.3, factual consistency rate)	95.9% ✎	▁▁▂▃▄▄█▆▃▂▂▁ beats 84/140	very easy · avg 94%	▇ 0.1%	blunt · maxed out · 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 70B Instruct · needs ~42.5 GB · from huggingface.co/lmstudio-community
chat in the terminal
llama cli -hf lmstudio-community/Meta-Llama-3-70B-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-3-70B-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 70B Instruct · needs ~42.5 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-3-70B-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 70B Instruct · needs ~42.83 GB · from ollama.com
chat in the terminal
ollama run llama3:70b

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 70B Instruct · needs ~39.7 GB, Mac only · from huggingface.co/mlx-community
chat in the terminal
mlx_lm.chat --model mlx-community/Meta-Llama-3-70B-Instruct-4bit
Or start an OpenAI-compatible server at http://localhost:8080/v1
mlx_lm.server --model mlx-community/Meta-Llama-3-70B-Instruct-4bit

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

$ how the score was worked out · EvalPlus / EvalPerf published… 24%, WildEval ZeroEval published r… 20%, Granite Code Models: A Family… 8% and 20 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	3	▇▇▇▇▇▇▇ 24%	Apache-2.0
WildEval ZeroEval published results	5	▇▇▇▇▇▇ 20%	Apache-2.0
Granite Code Models: A Family of Open Foundation Models for Code Intelligence — authors' own runs	3	▇▇ 8%	CC BY 4.0
LiveBench model judgments (HF livebench/model_judgment, 2024-10-22 revision) with the LiveBench question datasets	1	▇▇ 8%	Apache-2.0 (stated in the LiveBench datasheet)
AlpacaEval 1.0 and 2.0 leaderboards (tatsu-lab/alpaca_eval repository)	1	▇▇ 8%	Apache-2.0
Epoch AI — internal evaluations	3	▇▇ 6%	CC BY 4.0
Arena-Hard-Auto v0.1 leaderboard (repository README, pre-v2)	1	▇▇ 5%	Apache-2.0
BigCodeBench paper (Zhuo et al., ICLR 2025), arXiv 2406.15877v4	2	▇ 4%	CC BY 4.0
LiveCodeBench official website results (2024-12-26)	1	▇ 4%	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
WebApp1K models leaderboard (onekq-ai)	1	▇ 1%	MIT
Inception Arabic/Hindi leaderboards (AraGen-v3, HindiGen-v1, Arabic IFEval)	1	▇ 1%	Apache-2.0
Open Persian LLM Leaderboard v2 (Part DP AI and AUT NLP Lab)	1	▇ 1%	Apache-2.0
Knowledge Index of Noah's Ark (arXiv:2606.05104v2)	1	▇ 1%	CC BY 4.0
SciCode leaderboard (repository README)	1	▇ 0%	Apache-2.0
Berkeley Function Calling Leaderboard result archive	1	▇ 0%	Apache-2.0
Open Portuguese LLM Leaderboard (archived)	1	▇ 0%	Apache-2.0
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		= 49.6