whatmodel.sh / Llama 2 13B (base)

Llama 2 13B (base)

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

Llama 2 13B (base) is an open-weights language model from Meta, released July 2023. Its published size is 13B parameters and a 4k-token context window. Licence: llama2.

Catalogue identity notes

Pretrained (not chat) Llama 2 checkpoint; the chat models are separate identities (llama-2-*-chat). Ollama 'text' tags are the pretrained model.

34.0±0.5#608 of 719 LLMsfrom 53 benchmark results across 9 sources

$ how it compares

The LLMs ranked either side of Llama 2 13B (base), out of 828 with a score, then more from Meta.

#modelscorewhereRAMreleased
701StarCoder 15.5B BigCode35±1local~11GB2023-04
702Granite 3.1 3B-A800M Instruct IBM35±2local~3.6GB2024-12
703Llama 2 13B (base) Meta34±1local~9.3GB2023-07
704Code Llama 7B (base) Meta34±1local~5.5GB2023-08
705Nous-Hermes-13b Nous Research34±2local~9.8GB2023-06
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

coding▁▁▃▃▆██▇▆▄▃▁ #443 of 458 · 14% of its score · 3 tests

Beats Molmo 7B-D (0924), DeepSeek R1 Distill Qwen 1.5B, Tulu 2 7B here, though they rank higher overall. Trails Falcon3 1B Instruct, Granite 3.0 1B-A400M Instruct, StarCoderBase 3B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
HumanEval +218.9% ✎█···█·███··█ beats 1/5easy · avg 66%▇▇▇ 7.9%sharp · several runs
MBPP +235.4% ✎▄··▄·▄··█··▄ beats 1/5easy · avg 71%▇ 3.7%at its level · several runs
Tulu 2: CodexEval pass@1032.5% ✎▃▃▃·▆▃▃█▃▂▂▅ beats 2/25very easy · avg 86%▇ 2.2%sharp · at its level
math▁▃▃▅▆▅▄█▄▁▁▁ #375 of 398 · 7.8% of its score · 2 tests

Beats Command R (03-2024), Molmo 7B-D (0924), Mistral 7B Instruct v0.3 here, though they rank higher overall. Trails Gemma 2B (base), SmolLM2 1.7B Instruct, LFM2 350M here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
GSM8K +325.0% ✎▆▂▅█▅▅▃▆···▂ beats 7/25easy · avg 82%▇▇ 5.6%sharp · at its level · several runs
MATH +16.0% ✎█·█·█·██···█ beats 1/5hard · avg 30%▇ 2.2%several runs
other tests21 not in a task · 78% of its score
testresultvs the others in its rundifficultycountswhy
ARC Challenge +349.2% ✎▁·▁··▁▁▇▂█▆▃ beats 25/26easy · avg 85%▇▇▇▇▇▇▇▇▇▇▇ 27%sharp · at its level · several runs
WinoGrande +272.2% ✎▁····▁··▁▂█▄ beats 24/26easy · avg 89%▇▇▇ 7.7%at its level · several runs
Natural Questions (closed book) +232.2% ✎▁·▁·▁▂▁▂·▅▂█ beats 28/28medium · avg 58%▇▇▇ 6.6%several runs
BIG-Bench Hard +148.9% ✎▁·····▂█▂▅·▁ beats 9/25easy · avg 77%▇▇▇ 6.5%sharp · several runs
MMLU +352.0% ✎▁▁▃▃▄█▄▁█··▁ beats 12/25easy · avg 82%▇▇ 6.1%at its level · several runs
HellaSwag +279.4% ✎▃···▁▇▅█▇▃▁▁ beats 23/26very easy · avg 92%▇▇ 4.1%at its level · several runs
ARC Easy +277.6% ✎▁▁·▁▁▄▅▄█▅▁▃ beats 25/26easy · avg 88%▇ 3.7%at its level · several runs
BoolQ80.5% ✎▂▃▂▅▆▂▅▆█▂▃▂ beats 25/26easy · avg 91%▇ 3.3%at its level
13 more tests
testresultvs the others in its rundifficultycountswhy
TriviaQA +29.5% ✎▁··▁█▃▃▁▄▅▄█ beats 5/28very easy · avg 86%▇ 3.3%at its level · several runs
PIQA +279.1% ✎▁·▁··▁▁▃█▅█▂ beats 22/26easy · avg 88%▇ 2.4%blunt · at its level · several runs
OpenBookQA35.4% ✎█·▁▁▃·▁▃▅▇▃▄ beats 24/26medium · avg 53%▇ 1.8%typical
SciQ94.5% ✎▁·▁·▁·▃▅▃█▄▁ beats 22/26very easy · avg 98%▇ 1.4%maxed out
Social IQA42.8% ✎▁···▁▃█▃▁▁·▁ beats 20/26medium · avg 59%▇ 1.0%blunt
LAMBADA (OpenAI version), last-word accuracy76.8% ✎▁··▂▁▁▁▂▆█▅▃ beats 26/26very easy · avg 89%▇ 0.8%blunt
C-Eval37.3% ✎▁···▁▃▂▁▂▂▃█ beats 2/33medium · avg 62%▇ 0.6%noisy
Open PT-LLM Leaderboard average (9 Portuguese tasks)57.5% ✎▃▁▁▁▁▁▁▁▃█▅▁ beats 83/276medium · avg 73%▇ 0.5%at its level · noisy
Tulu 2: TyDiQA gold passage, 1-shot F156.5% ✎▂··▁▃▂▂▂█▅▁▂ beats 21/25medium · avg 61%▇ 0.5%blunt · at its level
EuroEval Life in the UK (English knowledge MC), test split, 5-shot, accuracy74.0% ✎▁▁▁▁▁▁▁▂▄█▄▃ beats 47/146easy · avg 83%▇ 0.4%at its level · noisy
HaluEval QA hallucination detection (accuracy)70.0% ✎█▁▁·▁▁·▁▁▁▁▃ beats 27/27medium · avg 70%▇ 0.3%sharp · noisy
TruthfulQA MC236.9% ✎▅▃█▆▁▂▅···▁▁ beats 5/31medium · avg 64%▇ 0.2%blunt · at its level
COPA85.0% ✎▅·▃·▃··█▃··▃ beats 6/8easy · avg 92%▇ 0.1%at its level · noisy

$ where to run it

llama.cpp · recommended

version

  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 2 13B (base) · needs ~7.9 GB · from huggingface.co/TheBloke
    chat in the terminal
    llama cli -hf TheBloke/Llama-2-13B-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf TheBloke/Llama-2-13B-GGUF:Q4_K_M

LM Studio

version

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

    download it (macOS / Windows / Linux) ↗

  2. Run Llama 2 13B (base) · needs ~7.9 GB · from huggingface.co/TheBloke

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/TheBloke/Llama-2-13B-GGUF@Q4_K_M -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 2 13B (base) · needs ~7.4 GB · from ollama.com
    chat in the terminal
    ollama run llama2:13b-text

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

$ how the score was worked out · Stability AI — original Stabl… 38%, Mixtral of Experts — base-mod… 22%, MiniCPM: Unveiling the Potent… 13% and 6 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
Stability AI — original StableLM zero-shot evaluation outputs10▇▇▇▇▇▇▇▇▇▇▇ 38%Apache-2.0
Mixtral of Experts — base-model evaluations12▇▇▇▇▇▇▇ 22%CC BY 4.0
MiniCPM: Unveiling the Potential of Small Language Models with Scalable Training Strategies (Hu et al.)7▇▇▇▇ 13%CC BY 4.0
Mistral 7B — authors' own runs12▇▇▇▇ 12%CC BY 4.0
Tulu 2 — base and instruction-tuned evaluations5▇▇▇ 8%CC BY 4.0
Hallucinations Leaderboard results (HF dataset hallucinations-leaderboard/results)4▇ 5%Apache-2.0
Open Portuguese LLM Leaderboard (archived)1▇ 0%Apache-2.0
EuroEval leaderboards (results collection), English generative leaderboard1▇ 0%MIT
From Test-Taking to Test-Making: Examining LLM Authoring of Commonsense Assessment Items (arXiv:2410.14897v1)1▇ 0%CC-BY-4.0
prior · pull toward the middle▇ 0%
weighted average= 34.0

$ 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