whatmodel.sh / Llama 4 Maverick (402B-A17B)

Llama 4 Maverick (402B-A17B)

+ contribute

Llama 4 Maverick (402B-A17B) is an open-weights language model from Meta, released April 2025. Its published size is 402B parameters (17B active per token) and a 1M-token context window. Licence: llama-4.

56.8±0.8#238 of 719 LLMsfrom 27 benchmark results across 20 sources

$ how it compares

The LLMs ranked either side of Llama 4 Maverick (402B-A17B), out of 828 with a score, then more from Meta.

#	model	score	where	RAM	released
282	GPT-4o (2024-05-13) OpenAI	57±0	api $7.50/M		2024-05
283	QED-Nano LM Provers	57±2	local	~4.1GB	2026-02
284	Llama 4 Maverick (402B-A17B) Meta	57±1	api $0.42/M +local	~65GB	2025-04
285	Claude 3.5 Sonnet (2024-06-20) Anthropic	57±1	comparison only		2024-06
286	Qwen2.5 Coder 32B Instruct Alibaba	57±1	api $0.74/M +local	~21GB	2024-11
	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▁▁▃▆█▆▅▅▅▄▃▁ #239 of 300 · 8.8% of its score · 5 tests

Beats DeepSeek R1 0528 Qwen3 8B here, though they rank higher overall. Trails Granite 4.0 1B, SmolLM3 3B, Phi-3.5-mini-instruct here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
BFCL V4 (December 2025 archive)	37.3% ✎	▁▁▁▂██▅▅▆▂▂▂ beats 42/77	hard · avg 32%	▇ 2.9%	at its level
Finance Agent Benchmark v1 (537 questions, class-balanced accuracy)	3.1% ✎	█·▃▃·█▅▅▅▅▃▅ beats 2/18	very hard · avg 5.8%	▇ 2.7%	sharp
BFCL v3	52.9% ✎	▁▁▁▁▁▃▅█▅▄▁▁ beats 49/114	medium · avg 52%	▇ 1.3%	blunt · at its level
Multi-SWE-bench, 7 languages (MagentLess scaffold)	6.1% ✎	▄···▄▄·▄▄·▄█ beats 6/7	very hard · avg 2.6%	▇ 1.1%	blunt · too hard for it
Galileo Agent Leaderboard v1 (tool selection quality)	74.1% ✎	▃▃▃▃▆▅▆▅█▂▅▂ beats 4/31	easy · avg 79%	▇ 0.9%	blunt

coding▁▁▁▃▆██▇▆▄▃▁ #215 of 458 · 17% of its score · 3 tests

Beats Qwen3.5 4B, Qwen3-VL 32B Instruct, Solar Open 100B (102B-A12B) here, though they rank higher overall. Trails QwQ-32B-Preview, Ministral 3 3B Reasoning 2512, DeepSeek-Coder-V2 Instruct (0724) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Aider polyglot (225 exercises, two attempts)	15.6% ✎	▂·▂▃▅▂▃█▆▄▂▂ beats 6/49	very hard · avg 8.5%	▇▇▇▇▇ 13%	sharp · at its level
WebApp1K-React pass@1 (onekq-ai)	76.9% ✎	▃▁▁█▁▄▅▃▄▇▁▄ beats 19/31	medium · avg 44%	▇ 2.4%	sharp · at its level · noisy
Multi-SWE-bench, 7 languages (MagentLess scaffold)	6.1% ✎	▄···▄▄·▄▄·▄█ beats 6/7	very hard · avg 2.6%	▇ 1.1%	blunt · too hard for it

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #190 of 393 · 15% of its score · 3 tests

Beats Seed-OSS 36B Instruct, Granite 4.2 8B, Qwen3-VL 30B-A3B Thinking here, though they rank higher overall. Trails LFM2.5 1.2B Instruct, LFM2.5 1.2B Thinking, Gemma 3 12B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Arena-Hard v2.0 hard prompts (style control, Gemini-2.5 judge)	17.2% ✎	▂·▃▅▅·▃█▆▂▂▂ beats 7/22	very hard · avg 12%	▇▇▇ 7.8%	sharp · at its level
LMArena (text)	1327.0 ✎	▁▁▁▂▃▃▃▆▆▇█▃ beats 165/379	avg 1268	▇▇ 5.9%	blunt
Arena-Hard v2.0 creative writing (GPT-4.1 + Gemini-2.5 judges)	10.5% ✎	▆▂▂▆▂▂▃█▂▃·▂ beats 2/22	hard · avg 19%	▇ 1.5%	sharp · at its level · noisy

math▁▂▃▅▆▅▄█▄▁▁▁ #195 of 398 · 20% of its score · 3 tests

Beats MiniMax-M2.1, Qwen3 235B-A22B (Apr 2025, hybrid), DeepSeek V3 (Dec '24) here, though they rank higher overall. Trails Phi-4-mini-reasoning 3.8B, DeepSeek R1 Distill Qwen 7B, OpenThinker2 7B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
MATH level 5	73.0% ✎	▂▂▅▆▄█▆▃▅▂▃▄ beats 62/92	hard · avg 32%	▇▇▇▇▇ 14%	sharp · at its level
OTIS mock AIME 2024–2025	20.6% ✎	▃▃▄▃▃▃▇▇█▃▂▅ beats 58/193	very hard · avg 8.5%	▇▇ 4.9%	sharp · at its level · noisy
MathNet-Solve-Test micro-average accuracy (6,400 olympiad problems)	14.7% ✎	▃▃···█▃·▃·▃▃ beats 0/8	hard · avg 15%	▇ 1.6%	typical

vision & documents▁▁▃▂▁▃▄▂▆▇█▄ #95 of 147 · 11% of its score · 4 tests

Beats DeepSeek V4 Flash Vision (Exp), Magistral Small 1.2, Qwen3-VL 30B-A3B Instruct here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
Arena vision (style controlled)	1145.2 ✎	▁▁▂▂▂▂▄▅▅▆██ beats 46/145	avg 1068	▇▇ 4.1%	blunt
CharXiv (reasoning questions)	55.0% ✎	▂▅▃▅▅▆▆▅▆█▆▂ beats 35/100	medium · avg 38%	▇ 2.6%	at its level · small run
CharXiv descriptive questions	87.7% ✎	▂▄▄▅▇▄█▇▇▂▁▂ beats 40/58	easy · avg 72%	▇ 2.2%	small run
VMMU (Vietnamese multimodal understanding and reasoning, mean accuracy over 7 tasks)	38.4% ✎	▅······▃·▅·█ beats 3/7	very hard · avg 36%	▇ 1.6%	at its level · noisy

web search▂▁▂▆▄█▃▃▁▁·▁ #54 of 80 * · 2.7% of its score · 1 test

Beats Kimi K2.6, GPT-5 Pro, DeepSeek V4 Flash (Preview, Apr 2026) here, though they rank higher overall. Trails GPT-4o mini (2024-07-18), Mistral Small 3.1 24B (2503), Llama 4 Scout (109B-A17B) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Finance Agent Benchmark v1 (537 questions, class-balanced accuracy)	3.1% ✎	█·▃▃·█▅▅▅▅▃▅ beats 2/18	very hard · avg 5.8%	▇ 2.7%	sharp

other tests10 not in a task · 29% of its score

test	result	vs the others in its run	difficulty	counts	why
GPQA Diamond	67.0% ✎	▁▁▁▁▂▄▆▄▅▇█▅ beats 101/208	hard · avg 40%	▇▇▇▇ 10%	sharp · at its level
KCL-Essay vanilla	25.6% ✎	▁···▁▃▃█▂▃▃▂ beats 6/23	hard · avg 18%	▇ 3.6%	at its level
EQ-Bench 3 (emotional intelligence, judge Elo)	1062.0 ✎	▁·▂▁▄▅▅▃█▄▁▁ beats 8/62	avg 1044	▇ 3.6%	typical
CPTUB average (SpeakLeash, 0-5 judge scale)	78.7% ✎	▁▂▂▂▃█▃▆▅▄▂▃ beats 77/95	easy · avg 66%	▇ 3.3%	typical
KINA (Knowledge Index of Noah's Ark) accuracy, 899 ten-option items	31.6% ✎	▁▁▂▁▃▄█▆▄▅▂▃ beats 24/39	very hard · avg 22%	▇ 2.9%	blunt
KCL-MCQA vanilla	43.3% ✎	▂···█▂▆▆██▆▂ beats 13/23	very hard · avg 30%	▇ 1.7%	typical
LongSocialBench (full-context, 4-option MCQ over discussion threads)	41.2% ✎	▆▂·▄█·▂▂·▄▄▂ beats 6/16	very hard · avg 36%	▇ 1.6%	blunt
DiagnosticIQ industrial maintenance MCQ (macro accuracy, 4 options)	63.3% ✎	▁··▃▁▁▁▅█▅▁▅ beats 15/24	medium · avg 52%	▇ 1.2%	blunt · at its level

2 more tests

test	result	vs the others in its run	difficulty	counts	why
Open PT-LLM Leaderboard average (9 Portuguese tasks)	80.1% ✎	▃▁▁▁▁▁▁▁▂█▅▁ beats 260/276	medium · avg 73%	▇ 1.1%	at its level · noisy
Vectara hallucination leaderboard (HHEM-2.3, factual consistency rate)	91.8% ✎	▁▂▁▂▅█▃▃▄▃▁▁ beats 62/98	very easy · avg 90%	▇ 0.0%	blunt

$ where to run it

Through an API

api · Amazon Bedrock · $0.42/M · $0.24 in / $0.97 out per 1M tokens · 1M context · docs.aws.amazon.com ↗ · price sourceUSD per 1M tokens, Amazon Bedrock on-demand Llama 4 Maverick 17B (Meta Llama 4 On-Demand table), US East (N. Virginia)/Ohio and US West (Oregon/N. California). Batch pricing excluded. Table cells are rendered from AWS's own price data file for the page (published 2026-09-25T06:55:02Z). Currency: USD. Amazon Bedrock official pricing page, observed 2026-09-25.

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 4 Maverick (402B-A17B) · needs ~121.1 GB · from huggingface.co/unsloth
chat in the terminal
llama cli -hf unsloth/Llama-4-Maverick-17B-128E-Instruct-GGUF:UD-Q2_K_XL
Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf unsloth/Llama-4-Maverick-17B-128E-Instruct-GGUF:UD-Q2_K_XL

LM Studio

version

Install LM Studio · Open the app once before using the lms command.
download it (macOS / Windows / Linux) ↗
Run Llama 4 Maverick (402B-A17B) · needs ~121.1 GB · from huggingface.co/unsloth
open in LM Studio ↗
Or download it in the terminal
lms get https://huggingface.co/unsloth/Llama-4-Maverick-17B-128E-Instruct-GGUF@UD-Q2_K_XL -y
Then run lms ls to see its name, and lms chat followed by that name

Ollama

Install Ollama
download it (macOS / Windows) ↗
Linux
curl -fsSL https://ollama.com/install.sh | sh
Run Llama 4 Maverick (402B-A17B) · needs ~241.7 GB · from ollama.com
chat in the terminal
ollama run llama4:maverick

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

$ how the score was worked out · Epoch AI — internal evaluatio… 29%, Aider polyglot benchmark runs 13%, LMArena leaderboard dataset 10% and 17 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
Epoch AI — internal evaluations	3	▇▇▇▇▇▇▇▇▇ 29%	CC BY 4.0
Aider polyglot benchmark runs	1	▇▇▇▇ 13%	Apache-2.0
LMArena leaderboard dataset	2	▇▇▇ 10%	CC BY 4.0
Arena-Hard-Auto v2.0-Preview leaderboard (repository README)	2	▇▇▇ 9%	Apache-2.0
Korean Canonical Legal Benchmark: Toward Knowledge-Independent Evaluation of LLMs’ Legal Reasoning Capabilities	2	▇▇ 5%	CC BY 4.0
Apriel-1.5-15b-Thinker (arXiv:2510.01141v1), Table 4 multimodal benchmarks	2	▇ 5%	CC-BY-4.0
Berkeley Function Calling Leaderboard result archive	2	▇ 4%	Apache-2.0
EQ-Bench 3 canonical leaderboard results	1	▇ 4%	MIT
SpeakLeash CPTUB (Complex Polish Text Understanding Benchmark)	1	▇ 3%	MIT
Knowledge Index of Noah's Ark (arXiv:2606.05104v2)	1	▇ 3%	CC BY 4.0
Finance Agent Benchmark (Vals AI, arXiv:2508.00828v1)	1	▇ 3%	CC-BY-4.0
WebApp1K models leaderboard (onekq-ai)	1	▇ 2%	MIT
MathNet: a Global Multimodal Benchmark for Mathematical Reasoning and Retrieval — authors' own model runs	1	▇ 2%	CC BY 4.0
VMMU: A Vietnamese Multitask Multimodal Understanding and Reasoning Benchmark (arXiv:2508.13680v5)	1	▇ 2%	CC-BY-4.0
LongSocialBench: Do Long-Context LLMs Understand Online Discussion Threads? (arXiv:2610.04118v1)	1	▇ 2%	CC BY 4.0
DiagnosticIQ: A Benchmark for LLM-Based Industrial Maintenance Action Recommendation from Symbolic Rules (arXiv:2605.08614v1)	1	▇ 1%	CC BY 4.0
Open Portuguese LLM Leaderboard (archived)	1	▇ 1%	Apache-2.0
Multi-SWE-bench experiments (ByteDance Seed baseline runs)	1	▇ 1%	Apache-2.0
Galileo Agent Leaderboard (v1 and v2 results)	1	▇ 1%	Apache-2.0
Vectara hallucination leaderboard (vectara/hallucination-leaderboard repository)	1	▇ 0%	Apache-2.0
prior · pull toward the middle		▇ 0%
weighted average		= 56.8