whatmodel.sh / Llama 3.1 405B Instruct

Llama 3.1 405B Instruct

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

Llama 3.1 405B Instruct is an open-weights language model from Meta, released July 2024. Its published size is 405B parameters and a 128k-token context window. Licence: llama-3.1.

54.9±0.5#268 of 719 LLMsfrom 36 benchmark results across 19 sources

$ how it compares

The LLMs ranked either side of Llama 3.1 405B Instruct, out of 828 with a score, then more from Meta.

#modelscorewhereRAMreleased
320Spark-X2.5 1.7B iFLYTEK (SparkLLM)55±1local~2.7GB2026-08
321GigaChat 3.5 Ultra Instruct (432B-A28B) Sber55±1local~457GB2026-07
322Llama 3.1 405B Instruct Meta55±1api $2.40/M +local~160GB2024-07
323Qwen2.5 32B Instruct Alibaba55±1local~21GB2024-09
324Command A Reasoning (08-2025) Cohere55±2local~34GB
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▁▁▃▆█▆▅▅▅▄▃▁ #203 of 300 · 12% of its score · 8 tests

Beats o1 (2024-12-17), o3-mini, DeepSeek R1 (Jan '25) here, though they rank higher overall. Trails SmolLM3 3B, Jamba 1.5 Mini, Mellum2 12B-A2.5B Instruct here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
SWE-bench Verified33.4% ✎▃···▃·▃·▃█·▃ beats 6/7hard · avg 19%▇ 3.0%sharp · at its level
MASSIVE-Agents AST (10k, zero-shot)25.6% ✎█····███·███ beats 5/6hard · avg 16%▇ 2.7%sharp · at its level
BrowserGym WebArena (AgentLab GenericAgent)24.0% ✎█··▄·▄█·▄··▄ beats 2/7hard · avg 18%▇ 1.6%typical
BrowserGym WorkArena-L1 (AgentLab GenericAgent)43.3% ✎█···█▄▄·▄█·█ beats 4/10hard · avg 29%▇ 1.5%at its level
BrowserGym MiniWoB (AgentLab GenericAgent)64.6% ✎▅····▅▅▃▃·▃█ beats 4/11medium · avg 55%▇ 0.9%blunt · at its level
τ-bench Airline26.0% ✎▁▁▃·▅▄▁█▇▁▁▇ beats 2/28hard · avg 23%▇ 0.7%at its level
BrowserGym WorkArena-L2 (AgentLab GenericAgent)7.2% ✎▃▅▃·▅▃▃··█·▃ beats 4/11very hard · avg 3.9%▇ 0.7%sharp · noisy
τ-bench Retail29.1% ✎▁····▃·▂▇█▂▆ beats 2/28hard · avg 29%▇ 0.6%at its level · noisy
coding▁▁▁▃▆██▇▆▄▃▁ #214 of 458 · 29% of its score · 10 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.

testresultvs the others in its rundifficultycountswhy
LiveCodeBench official (Jan 1–Nov 1, 2024)37.3% ✎▁▁·▁▂▄▆█▅▂·▂ beats 22/31hard · avg 26%▇▇▇ 6.8%sharp · at its level
CRUXEval (ZeroEval, 800 problems)73.5% ✎▁▂▆▃▆█▆▃▂·▁▁ beats 32/47medium · avg 61%▇▇ 5.8%sharp · at its level
Aider code editing (133 Exercism Python exercises, two attempts)66.2% ✎▁·▁▂▄▁▆▄█▆▂▂ beats 39/62medium · avg 54%▇▇ 4.4%at its level
SWE-bench Verified33.4% ✎▃···▃·▃·▃█·▃ beats 6/7hard · avg 19%▇ 3.0%sharp · at its level
Aider polyglot (225 exercises, two attempts)13.8% ✎█▄··█·█····▄ beats 5/7very hard · avg 6.7%▇ 2.9%sharp · noisy
LiveCodeBench v5 (Oct 2024–Feb 2025)29.3% ✎▄▄·▄·····█▄█ beats 5/7hard · avg 19%▇ 1.9%at its level
SciCode main problems (official leaderboard)1.5% ✎▃···█··▂·▃·▂ beats 10/17very hard · avg 1.1%▇ 1.3%sharp · too hard for it
MBPP+87.0% ✎██·██···████ beats 5/7easy · avg 79%▇ 1.2%typical
2 more tests
testresultvs the others in its rundifficultycountswhy
WebApp1K-React pass@1 (onekq-ai)30.2% ✎▃▁▁█▁▄▅▃▄▇▁▄ beats 5/31medium · avg 44%▇ 1.2%sharp · at its level · noisy
BIRD-SQL (dev)59.4% ✎▄··▄·▄·█··▄█ beats 6/7medium · avg 49%▇ 1.0%at its level
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #185 of 393 · 10% of its score · 3 tests

Beats Seed-OSS 36B Instruct, Nemotron-Cascade-14B-Thinking, Granite 4.2 8B 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
Arena-Hard (v0.1, GPT-4-1106 judge)69.3% ✎▂▁▁▅█▅▄▃▃▅▃▁ beats 45/60medium · avg 52%▇▇ 5.8%sharp · at its level
LMArena (text)1335.4 ✎▁▁▁▂▃▃▃▆▆▇█▃ beats 171/379avg 1268▇ 2.6%blunt
Arabic IFEval prompt-level (Inception)50.7% ✎▁·▁▁▂▁▄▄▆▄█▄ beats 27/38medium · avg 49%▇ 2.1%sharp · at its level
math▁▂▃▅▆▅▄█▄▁▁▁ #247 of 398 · 12% of its score · 4 tests

Beats GPT-4o (2024-11-20), GPT-4 Turbo (2024-04-09), Molmo 2 8B here, though they rank higher overall. Trails DeepSeek R1 Distill Qwen 1.5B, Tulu 2 DPO 70B, OpenThinker 7B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
MATH level 549.8% ✎▂▂▅▆▄█▆▃▅▂▃▄ beats 42/92hard · avg 32%▇▇▇ 6.4%sharp · at its level
AIME 202420.0% ✎▁▁▁▂▁▃▁▅███▃ beats 11/85hard · avg 16%▇ 2.4%sharp · at its level · small run
GSM8K (ZeroEval, zero-shot)96.0% ✎▁▄·▁▄█▅▅▅███ beats 40/43very easy · avg 89%▇ 1.8%sharp · maxed out
OTIS mock AIME 2024–20259.7% ✎▃▃▄▃▃▃▇▇█▃▂▅ beats 50/193very hard · avg 8.5%▇ 1.8%sharp · noisy
other tests12 not in a task · 39% of its score
testresultvs the others in its rundifficultycountswhy
LiveBench 2024-07-2654.3% ✎▁▁▁▂▂▃▂█▆▃▅▃ beats 58/61medium · avg 41%▇▇▇▇▇ 11%at its level
MMLU-Redux (ZeroEval, 2,778 questions)84.4% ✎▁·▁▁▄▄▄▅█▄▂▁ beats 42/53easy · avg 77%▇▇▇▇ 9.0%typical
ZebraLogic (ZeroEval, 1,000 puzzles)25.0% ✎▁▁▃▆██▁▁·▁▁▁ beats 38/57hard · avg 18%▇▇ 5.6%sharp · at its level
GPQA Diamond50.9% ✎▁▁▁▁▂▄▆▄▅▇█▅ beats 71/208hard · avg 40%▇▇ 4.2%sharp · at its level
RepoQA (32K, threshold 0.8)90.4% ✎▄▄▄··▄▄··▄·█ beats 5/7easy · avg 81%▇ 1.6%typical
EQ-Bench 3 (emotional intelligence, judge Elo)1058.4 ✎▁·▂▁▄▅▅▃█▄▁▁ beats 6/62avg 1044▇ 1.6%typical
BigCodeBench (Instruct)46.2% ✎▄··▄▄···▄▄█▄ beats 6/7medium · avg 36%▇ 1.3%at its level
Polish EQ-Bench v2 (SpeakLeash)77.2% ✎▁··▁▁▂▂▆▅▆█▆ beats 67/70medium · avg 64%▇ 1.2%at its level
4 more tests
testresultvs the others in its rundifficultycountswhy
KINA (Knowledge Index of Noah's Ark) accuracy, 899 ten-option items29.6% ✎▁▁▂▁▃▄█▆▄▅▂▃ beats 19/39very hard · avg 22%▇ 1.2%blunt
Judgemark v2 (LLM-as-judge for creative writing)60.1% ✎▃▃··▃▅▅···█▃ beats 4/10medium · avg 55%▇ 0.9%at its level
DiagnosticIQ industrial maintenance MCQ (macro accuracy, 4 options)56.6% ✎▁··▃▁▁▃▅█▅▁▅ beats 5/24medium · avg 52%▇ 0.5%blunt · at its level
Open PT-LLM Leaderboard average (9 Portuguese tasks)82.1% ✎▃▁▁▁▁▁▁▁▂█▅▃ beats 267/276medium · avg 73%▇ 0.5%at its level · noisy

$ where to run it

Through an API

api · Amazon Bedrock · $2.40/M · $2.4 in / $2.4 out per 1M tokens · 128k context · docs.aws.amazon.com ↗ · price sourceUSD per 1M tokens, Amazon Bedrock on-demand Llama 3.1 Instruct (405B), standard (not latency-optimized) row, offered in US West (Oregon) and US East (Ohio). Batch .20/.20 and the latency-optimized row 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

  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 3.1 405B Instruct · needs ~151.2 GB · from huggingface.co/MaziyarPanahi
    chat in the terminal
    llama cli -hf MaziyarPanahi/Meta-Llama-3.1-405B-Instruct-GGUF:Q2_K
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf MaziyarPanahi/Meta-Llama-3.1-405B-Instruct-GGUF:Q2_K

LM Studio

version

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

    download it (macOS / Windows / Linux) ↗

  2. Run Llama 3.1 405B Instruct · needs ~151.2 GB · from huggingface.co/MaziyarPanahi

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/MaziyarPanahi/Meta-Llama-3.1-405B-Instruct-GGUF@Q2_K -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 3.1 405B Instruct · needs ~243.5 GB · from ollama.com
    chat in the terminal
    ollama run llama3.1:405b

MLX (Mac)

  1. Install MLX (Mac) · Apple silicon Macs only, text only.
    with uv
    uv tool install mlx-lm
    or
    pip install mlx-lm
  2. Run Llama 3.1 405B Instruct · needs ~431.2 GB, Mac only · from huggingface.co/mlx-community
    chat in the terminal
    mlx_lm.chat --model mlx-community/Llama-3.1-405B-Instruct-8bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model mlx-community/Llama-3.1-405B-Instruct-8bit

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

$ how the score was worked out · WildEval ZeroEval published r… 22%, Cohere — Command A technical… 17%, Epoch AI — internal evaluatio… 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

sourceresultsshare of the scorelicence
WildEval ZeroEval published results4▇▇▇▇▇▇▇ 22%Apache-2.0
Cohere — Command A technical report10▇▇▇▇▇ 17%CC BY 4.0
Epoch AI — internal evaluations3▇▇▇▇ 12%CC BY 4.0
LiveBench model judgments (HF livebench/model_judgment, 2024-10-22 revision) with the LiveBench question datasets1▇▇▇ 11%Apache-2.0 (stated in the LiveBench datasheet)
LiveCodeBench official website results (2024-12-26)1▇▇ 7%CC-BY-SA-4.0
Arena-Hard-Auto v0.1 leaderboard (repository README, pre-v2)1▇▇ 6%Apache-2.0
ServiceNow BrowserGym leaderboard (AgentLab GenericAgent runs)4▇ 5%MIT
Aider code editing benchmark runs (old edit leaderboard)1▇ 4%Apache-2.0
MASSIVE-Agents — Kulkarni et al. (2025)1▇ 3%CC BY 4.0
LMArena leaderboard dataset1▇ 3%CC BY 4.0
Inception Arabic/Hindi leaderboards (AraGen-v3, HindiGen-v1, Arabic IFEval)1▇ 2%Apache-2.0
EQ-Bench 3 canonical leaderboard results1▇ 2%MIT
SciCode leaderboard (repository README)1▇ 1%Apache-2.0
SpeakLeash Polish EQ-Bench leaderboard1▇ 1%Apache-2.0
Knowledge Index of Noah's Ark (arXiv:2606.05104v2)1▇ 1%CC BY 4.0
WebApp1K models leaderboard (onekq-ai)1▇ 1%MIT
Judgemark v2 results (EQ-Bench)1▇ 1%MIT
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▇ 0%Apache-2.0
prior · pull toward the middle▇ 0%
weighted average= 54.9

$ 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