whatmodel.sh / Llama 4 Scout (109B-A17B)

Llama 4 Scout (109B-A17B)

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Llama 4 Scout (109B-A17B) is an open-weights language model from Meta, released April 2025. Its published size is 109B parameters (17B active per token) and a 328k-token context window. Licence: llama-4.

54.3±1.0#280 of 719 LLMsfrom 19 benchmark results across 14 sources

$ how it compares

The LLMs ranked either side of Llama 4 Scout (109B-A17B), out of 828 with a score, then more from Meta.

#modelscorewhereRAMreleased
332Falcon-H1 34B Instruct TII54±3local~23GB2025-05
333Llama 3.3 70B Instruct Meta54±1api $0.72/M +local~24GB2024-12
334Llama 4 Scout (109B-A17B) Meta54±1api $0.15/M +local~36GB2025-04
335Nemotron 3 Nano 4B NVIDIA54±1local~3.8GB2026-03
336Granite 4.1 8B IBM54±1api $0.11/M +local~7GB2026-04
more from Meta
24Muse Spark 1.3 Meta77±1api $2.00/M2026-09
40Muse Spark 1.2 Meta75±1api $2.00/M2026-08
43Muse Spark 1.1 Meta75±1api $2.00/M2026-07

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #271 of 300 · 11% of its score · 4 tests

Beats Magistral Small 1.0 (2506) here, though they rank higher overall. Trails Granite 4.0 H 350M, Qwen3 0.6B, Qwen2.5 1.5B Instruct here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
BFCL V4 (December 2025 archive)28.1% ✎▁▁▁▂██▅▅▆▂▂▂ beats 26/77hard · avg 32%▇▇ 4.4%at its level
Finance Agent Benchmark v1 (537 questions, class-balanced accuracy)5.8% ✎█·▃▃·█▅▅▅▅▃▅ beats 4/18very hard · avg 5.8%▇ 3.4%sharp
BFCL v345.4% ✎▁▁▁▁▁▃▅█▅▄▁▁ beats 24/114medium · avg 52%▇ 2.0%blunt · at its level
Galileo Agent Leaderboard v1 (tool selection quality)62.9% ✎▃▃▃▃▆▅▆▅█▂▅▂ beats 0/31easy · avg 79%▇ 1.5%blunt
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #193 of 393 * · 9.3% of its score · 1 test

Beats Granite 4.2 8B, Qwen3-VL 30B-A3B Thinking, Mistral Small 4 (119B-A6.5B) 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.

testresultvs the others in its rundifficultycountswhy
LMArena (text)1321.6 ✎▁▁▁▂▃▃▃▆▆▇█▃ beats 154/379avg 1268▇▇▇▇ 9.3%blunt
math▁▂▃▅▆▅▄█▄▁▁▁ #222 of 398 * · 29% of its score · 2 tests

Beats GPT-4o (2024-08-06), Command A (03-2025), GPT-4o (2024-11-20) here, though they rank higher overall. Trails Olmo 3 7B Instruct, DeepSeek R1 Distill Llama 8B, Phi-4-mini-reasoning 3.8B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
MATH level 562.3% ✎▂▂▅▆▄█▆▃▅▂▃▄ beats 52/92hard · avg 32%▇▇▇▇▇▇▇▇▇ 23%sharp · at its level
OTIS mock AIME 2024–20257.8% ✎▃▃▄▃▃▃▇▇█▃▂▅ beats 45/193very hard · avg 8.5%▇▇ 5.9%sharp · noisy
vision & documents▁▁▃▂▁▃▄▂▆▇█▄ #100 of 147 * · 8.6% of its score · 2 tests

Beats DeepSeek V4 Flash Vision (Exp), GPT-4o (2024-08-06), Qwen3-VL 4B Thinking here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
Arena vision (style controlled)1127.4 ✎▁▁▂▂▂▂▄▅▅▆██ beats 40/145avg 1068▇▇▇ 6.4%blunt
VMMU (Vietnamese multimodal understanding and reasoning, mean accuracy over 7 tasks)49.1% ✎▅······▃·▅·█ beats 5/7very hard · avg 36%▇ 2.2%noisy
web search▂▁▂▆▄█▃▃▁▁·▁ #33 of 80 * · 3.4% of its score · 1 test

Beats Gemini 3.1 Pro (Preview), GPT-5.3-Codex, Qwen3.7 Max here, though they rank higher overall. Trails GPT-4o mini (2024-07-18), Mistral Small 3.1 24B (2503) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Finance Agent Benchmark v1 (537 questions, class-balanced accuracy)5.8% ✎█·▃▃·█▅▅▅▅▃▅ beats 4/18very hard · avg 5.8%▇ 3.4%sharp
other tests10 not in a task · 41% of its score
testresultvs the others in its rundifficultycountswhy
GPQA Diamond51.8% ✎▁▁▁▁▂▄▆▄▅▇█▅ beats 73/208hard · avg 40%▇▇▇▇▇▇ 15%sharp · at its level
EQ-Bench 3 (emotional intelligence, judge Elo)993.6 ✎▁·▃▁▄▅▅▃█▄▁▁ beats 3/62avg 1044▇▇ 5.7%typical
CPTUB average (SpeakLeash, 0-5 judge scale)75.0% ✎▁▂▂▂▃█▃▆▅▄▂▃ beats 69/95easy · avg 66%▇▇ 5.7%at its level
BRIDGE clinical text, zero-shot average (87 tasks)35.1% ✎▁▁▁▁▁▂▂▃▄▆█▃ beats 45/97hard · avg 32%▇▇ 4.9%blunt · at its level
Open Persian LLM Leaderboard v2 average (21 tasks)69.0% ✎▂▄▂▃█▄▃▆▃▅▂▂ beats 70/81medium · avg 58%▇▇ 3.8%blunt · at its level
LongSocialBench (full-context, 4-option MCQ over discussion threads)40.2% ✎▆▂·▄█·▂▂·▄▄▂ beats 5/16very hard · avg 36%▇ 2.3%blunt
Open PT-LLM Leaderboard average (9 Portuguese tasks)78.7% ✎▃▁▁▁▁▁▁▁▂█▅▁ beats 246/276medium · avg 73%▇ 1.8%at its level · noisy
EuroEval HellaSwag (English), test split, 5-shot generative MC, accuracy86.4% ✎▃·▁▁▂▁▂▅█▇▄▁ beats 86/100medium · avg 68%▇ 1.5%at its level · noisy
2 more tests
testresultvs the others in its rundifficultycountswhy
EuroEval Life in the UK (English knowledge MC), test split, 5-shot, accuracy91.0% ✎▁▁▁▁▁▁▁▂▄█▄▃ beats 120/146easy · avg 83%▇ 0.8%noisy
Vectara hallucination leaderboard (HHEM-2.3, factual consistency rate)92.3% ✎▁▂▁▂▅█▂▃▄▃▁▁ beats 64/98very easy · avg 90%▇ 0.0%blunt

$ where to run it

Through an API

api · DeepInfra · $0.15/M · $0.1 in / $0.3 out per 1M tokens · 328k context · deepinfra.com ↗ · price sourceUSD per 1M text tokens, standard uncached input and output. Product: Serverless text generation. API model: meta-llama/Llama-4-Scout-17B-16E-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

  1. 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 ↗

  2. Run Llama 4 Scout (109B-A17B) · needs ~33.2 GB · from huggingface.co/unsloth
    chat in the terminal
    llama cli -hf unsloth/Llama-4-Scout-17B-16E-Instruct-GGUF:UD-Q2_K_XL
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf unsloth/Llama-4-Scout-17B-16E-Instruct-GGUF:UD-Q2_K_XL

LM Studio

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run Llama 4 Scout (109B-A17B) · needs ~33.2 GB · from huggingface.co/unsloth

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/unsloth/Llama-4-Scout-17B-16E-Instruct-GGUF@UD-Q2_K_XL -y
  4. Then run lms ls to see its name, and lms chat followed by that name

Ollama

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run Llama 4 Scout (109B-A17B) · needs ~65.9 GB · from ollama.com
    chat in the terminal
    ollama run llama4:scout

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

$ how the score was worked out · Epoch AI — internal evaluatio… 44%, LMArena leaderboard dataset 16%, Berkeley Function Calling Lea… 6% and 11 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

sourceresultsshare of the scorelicence
Epoch AI — internal evaluations3▇▇▇▇▇▇▇▇▇▇▇▇▇ 44%CC BY 4.0
LMArena leaderboard dataset2▇▇▇▇▇ 16%CC BY 4.0
Berkeley Function Calling Leaderboard result archive2▇▇ 6%Apache-2.0
EQ-Bench 3 canonical leaderboard results1▇▇ 6%MIT
SpeakLeash CPTUB (Complex Polish Text Understanding Benchmark)1▇▇ 6%MIT
BRIDGE Medical Leaderboard (Harvard Medical School / Brigham and Women's YLab)1▇ 5%MIT
Open Persian LLM Leaderboard v2 (Part DP AI and AUT NLP Lab)1▇ 4%Apache-2.0
Finance Agent Benchmark (Vals AI, arXiv:2508.00828v1)1▇ 3%CC-BY-4.0
EuroEval leaderboards (results collection), English generative leaderboard2▇ 2%MIT
LongSocialBench: Do Long-Context LLMs Understand Online Discussion Threads? (arXiv:2610.04118v1)1▇ 2%CC BY 4.0
VMMU: A Vietnamese Multitask Multimodal Understanding and Reasoning Benchmark (arXiv:2508.13680v5)1▇ 2%CC-BY-4.0
Open Portuguese LLM Leaderboard (archived)1▇ 2%Apache-2.0
Galileo Agent Leaderboard (v1 and v2 results)1▇ 2%Apache-2.0
Vectara hallucination leaderboard (vectara/hallucination-leaderboard repository)1▇ 0%Apache-2.0
prior · pull toward the middle▇ 0%
weighted average= 54.3
$ published speed measurements · 2

Reference results for the hardware, engine and settings below. They aren't verified speeds of the downloads above or predictions for your machine.

AMD Ryzen AI Max+ 395 (Strix Halo), Radeon 8060S iGPU; Beelink GTR9 Pro 128GBQ4_K_M18.32 tok/sllama.cpp b8933, Vulkan (RADV) · llama-bench pp512/tg128 as tabulated (tg_tps column = tg128 generation); exact flags not retained in a raw CSV for this historical row

Historical row sourced by the guide to its own BENCHMARKS.md/README.md; no raw CSV retained. Row date 2026-04-26, status measured-local, kernel 6.19.4, driver stack 26.0.2-26.0.6. Author note: 109B parameters on one mini PC

source ↗
AMD Ryzen AI Max+ 395 (Strix Halo), Radeon 8060S; pre-production Framework Desktop 128GB LPDDR5x-8000UD-Q4_K_XL19.78 tok/sllama.cpp 83bc2f288 (commit from system_info.json), backend vulkan · llama-bench tg128 (128 generated tokens, empty context); backend variant 'llama.cpp-vulkan'; flags: defaults; mmap disabled per the guide; value = bold best tg128 in this model's Benchmark Results table

Author's own sweep; README value 19.784728 rounded to 2 decimals. Kernel 6.16.0-1-mainline; system_info timestamp 2025-08-05T09:07:01.852368Z; CPU string 'AMD RYZEN AI MAX+ 395 w/ Radeon 8060S'. Pre-production hardware and BIOS per the README.

source ↗

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json