Instructions to use YTan2000/Laguna-XS-2.1-TQ3_4S with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use YTan2000/Laguna-XS-2.1-TQ3_4S with llama.cpp:
Install (macOS, Linux)
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf YTan2000/Laguna-XS-2.1-TQ3_4S # Run inference directly in the terminal: llama cli -hf YTan2000/Laguna-XS-2.1-TQ3_4S
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf YTan2000/Laguna-XS-2.1-TQ3_4S # Run inference directly in the terminal: llama cli -hf YTan2000/Laguna-XS-2.1-TQ3_4S
Use pre-built binary
# Download pre-built binary from: # https://github.com/ggerganov/llama.cpp/releases # Start a local OpenAI-compatible server with a web UI: ./llama-server -hf YTan2000/Laguna-XS-2.1-TQ3_4S # Run inference directly in the terminal: ./llama-cli -hf YTan2000/Laguna-XS-2.1-TQ3_4S
Build from source code
git clone https://github.com/ggerganov/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-server llama-cli # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama-server -hf YTan2000/Laguna-XS-2.1-TQ3_4S # Run inference directly in the terminal: ./build/bin/llama-cli -hf YTan2000/Laguna-XS-2.1-TQ3_4S
Use Docker
docker model run hf.co/YTan2000/Laguna-XS-2.1-TQ3_4S
- LM Studio
- Jan
- vLLM
How to use YTan2000/Laguna-XS-2.1-TQ3_4S with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "YTan2000/Laguna-XS-2.1-TQ3_4S" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "YTan2000/Laguna-XS-2.1-TQ3_4S", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/YTan2000/Laguna-XS-2.1-TQ3_4S
- Ollama
How to use YTan2000/Laguna-XS-2.1-TQ3_4S with Ollama:
ollama run hf.co/YTan2000/Laguna-XS-2.1-TQ3_4S
- Unsloth Studio
How to use YTan2000/Laguna-XS-2.1-TQ3_4S with Unsloth Studio:
Install Unsloth Studio (macOS, Linux, WSL)
curl -fsSL https://unsloth.ai/install.sh | sh # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for YTan2000/Laguna-XS-2.1-TQ3_4S to start chatting
Install Unsloth Studio (Windows)
irm https://unsloth.ai/install.ps1 | iex # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for YTan2000/Laguna-XS-2.1-TQ3_4S to start chatting
Using HuggingFace Spaces for Unsloth
# No setup required # Open https://huggingface.co/spaces/unsloth/studio in your browser # Search for YTan2000/Laguna-XS-2.1-TQ3_4S to start chatting
- Pi
How to use YTan2000/Laguna-XS-2.1-TQ3_4S with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf YTan2000/Laguna-XS-2.1-TQ3_4S
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "llama-cpp": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "YTan2000/Laguna-XS-2.1-TQ3_4S" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Docker Model Runner
How to use YTan2000/Laguna-XS-2.1-TQ3_4S with Docker Model Runner:
docker model run hf.co/YTan2000/Laguna-XS-2.1-TQ3_4S
- Lemonade
How to use YTan2000/Laguna-XS-2.1-TQ3_4S with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull YTan2000/Laguna-XS-2.1-TQ3_4S
Run and chat with the model
lemonade run user.Laguna-XS-2.1-TQ3_4S-{{QUANT_TAG}}List all available models
lemonade list
- Hermes Agent
How to use YTan2000/Laguna-XS-2.1-TQ3_4S with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf YTan2000/Laguna-XS-2.1-TQ3_4S
Configure Hermes
# Install Hermes: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash hermes setup # Point Hermes at the local server: hermes config set model.provider custom hermes config set model.base_url http://127.0.0.1:8080/v1 hermes config set model.default YTan2000/Laguna-XS-2.1-TQ3_4S
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use YTan2000/Laguna-XS-2.1-TQ3_4S with OpenClaw:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf YTan2000/Laguna-XS-2.1-TQ3_4S
Configure OpenClaw
# Install OpenClaw: npm install -g openclaw@latest # Register the local server and set it as the default model: openclaw onboard --non-interactive --mode local \ --auth-choice custom-api-key \ --custom-base-url http://127.0.0.1:8080/v1 \ --custom-model-id "YTan2000/Laguna-XS-2.1-TQ3_4S" \ --custom-provider-id llama-cpp \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
Upload four_way_compare.html with huggingface_hub
Browse files- four_way_compare.html +225 -0
four_way_compare.html
ADDED
|
@@ -0,0 +1,225 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
<!doctype html>
|
| 2 |
+
<html lang="en">
|
| 3 |
+
<head>
|
| 4 |
+
<meta charset="utf-8">
|
| 5 |
+
<meta name="viewport" content="width=1600, initial-scale=1">
|
| 6 |
+
<title>Laguna XS 2.1-TQ3_4S — 196 tok/s on a single RTX 3090</title>
|
| 7 |
+
<link rel="preconnect" href="https://fonts.googleapis.com">
|
| 8 |
+
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
|
| 9 |
+
<link href="https://fonts.googleapis.com/css2?family=Chakra+Petch:wght@500;600;700&family=IBM+Plex+Mono:wght@400;500;600;700&family=IBM+Plex+Sans:wght@400;500;600&display=swap" rel="stylesheet">
|
| 10 |
+
<style>
|
| 11 |
+
:root{
|
| 12 |
+
color-scheme: dark;
|
| 13 |
+
--bg0:#0a0f1d; --bg1:#0d1528;
|
| 14 |
+
--panel:#101a30; --line:#22304e; --line2:#182338;
|
| 15 |
+
--text:#eaf1fd; --muted:#8ea3c8; --dim:#5f7396;
|
| 16 |
+
--ls:#2fe6c8; /* Laguna S teal */
|
| 17 |
+
--lx:#4db8ff; /* Laguna XS azure */
|
| 18 |
+
--q27:#ffb545; /* 27B amber */
|
| 19 |
+
--q9:#a8e94c; /* 9B lime */
|
| 20 |
+
--orn:#c792ff; /* Ornith 35B violet */
|
| 21 |
+
--disp:"Chakra Petch",ui-sans-serif,system-ui,sans-serif;
|
| 22 |
+
--mono:"IBM Plex Mono",ui-monospace,SFMono-Regular,Menlo,monospace;
|
| 23 |
+
--body:"IBM Plex Sans",ui-sans-serif,system-ui,sans-serif;
|
| 24 |
+
}
|
| 25 |
+
*{box-sizing:border-box; margin:0; padding:0}
|
| 26 |
+
html,body{width:1600px; height:900px; overflow:hidden}
|
| 27 |
+
body{
|
| 28 |
+
color:var(--text); font-family:var(--body);
|
| 29 |
+
background:
|
| 30 |
+
radial-gradient(52rem 30rem at 8% -6%, rgba(47,230,200,.12), transparent 60%),
|
| 31 |
+
radial-gradient(46rem 28rem at 92% 0%, rgba(77,184,255,.10), transparent 60%),
|
| 32 |
+
radial-gradient(42rem 30rem at 70% 104%, rgba(199,146,255,.08), transparent 60%),
|
| 33 |
+
linear-gradient(160deg, var(--bg0) 0%, var(--bg1) 52%, #080d18 100%);
|
| 34 |
+
}
|
| 35 |
+
body::before{
|
| 36 |
+
content:""; position:absolute; inset:0; pointer-events:none;
|
| 37 |
+
background:
|
| 38 |
+
repeating-linear-gradient(0deg, rgba(140,170,220,.05) 0 1px, transparent 1px 56px),
|
| 39 |
+
repeating-linear-gradient(90deg, rgba(140,170,220,.04) 0 1px, transparent 1px 56px);
|
| 40 |
+
mask-image:radial-gradient(90rem 60rem at 50% 18%, #000 30%, transparent 92%);
|
| 41 |
+
}
|
| 42 |
+
.stage{position:relative; width:1600px; height:900px; padding:46px 64px 34px; display:flex; flex-direction:column}
|
| 43 |
+
|
| 44 |
+
/* ---------- header ---------- */
|
| 45 |
+
header{display:flex; justify-content:space-between; align-items:flex-end; gap:32px;
|
| 46 |
+
border-bottom:2px solid var(--line); padding-bottom:20px; margin-bottom:18px}
|
| 47 |
+
.eyebrow{font-family:var(--mono); font-size:15px; font-weight:600; letter-spacing:.26em; color:var(--ls);
|
| 48 |
+
text-transform:uppercase; display:flex; align-items:center; gap:12px; margin-bottom:16px;
|
| 49 |
+
text-shadow:0 0 18px rgba(47,230,200,.5)}
|
| 50 |
+
.eyebrow::before{content:""; width:34px; height:2px; background:var(--ls); box-shadow:0 0 10px rgba(47,230,200,.7)}
|
| 51 |
+
h1{font-family:var(--disp); font-weight:700; text-transform:uppercase;
|
| 52 |
+
font-size:72px; line-height:.92; letter-spacing:-.015em; color:#f6faff;
|
| 53 |
+
text-shadow:0 2px 30px rgba(77,184,255,.25)}
|
| 54 |
+
h1 .acc{color:var(--ls); text-shadow:0 0 26px rgba(47,230,200,.55)}
|
| 55 |
+
h1 .acc2{display:block; font-size:38px; letter-spacing:.01em; color:var(--lx);
|
| 56 |
+
text-shadow:0 0 22px rgba(77,184,255,.5); margin-top:8px}
|
| 57 |
+
.marquee{text-align:right; flex-shrink:0}
|
| 58 |
+
.marquee .big{font-family:var(--mono); font-weight:700; font-size:84px; line-height:.9; color:var(--lx);
|
| 59 |
+
text-shadow:0 0 38px rgba(77,184,255,.6)}
|
| 60 |
+
.marquee .big small{font-size:23px; font-weight:600; color:var(--muted); letter-spacing:.08em; margin-left:6px}
|
| 61 |
+
.marquee .mlabel{font-family:var(--mono); font-size:12px; letter-spacing:.2em; color:var(--dim);
|
| 62 |
+
text-transform:uppercase; margin-top:6px}
|
| 63 |
+
.marquee .mlabel b{color:var(--lx); font-weight:700}
|
| 64 |
+
.marquee .big2{font-family:var(--mono); font-weight:700; font-size:46px; line-height:.9; color:var(--ls);
|
| 65 |
+
text-shadow:0 0 32px rgba(47,230,200,.5); margin-top:16px}
|
| 66 |
+
.marquee .big2 small{font-size:16px; font-weight:600; color:var(--muted); letter-spacing:.08em; margin-left:5px}
|
| 67 |
+
.marquee .mlabel2{font-family:var(--mono); font-size:12px; letter-spacing:.2em; color:var(--dim);
|
| 68 |
+
text-transform:uppercase; margin-top:6px}
|
| 69 |
+
.marquee .mlabel2 b{color:var(--ls); font-weight:700}
|
| 70 |
+
|
| 71 |
+
/* headline stat strip */
|
| 72 |
+
.headstats{display:flex; gap:30px; margin-top:18px}
|
| 73 |
+
.hstat{display:flex; flex-direction:column; gap:3px; padding-left:14px; border-left:2px solid var(--line);
|
| 74 |
+
transition:border-color .25s, transform .25s}
|
| 75 |
+
.hstat:hover{border-color:var(--lx); transform:translateY(-2px)}
|
| 76 |
+
.hstat b{font-family:var(--mono); font-weight:700; font-size:22px; color:var(--text); line-height:1}
|
| 77 |
+
.hstat span{font-family:var(--mono); font-size:10.5px; font-weight:600; letter-spacing:.18em; text-transform:uppercase; color:var(--muted)}
|
| 78 |
+
.hstat.hx b{color:var(--lx); text-shadow:0 0 14px rgba(77,184,255,.4)}
|
| 79 |
+
.hstat.hs b{color:var(--ls); text-shadow:0 0 14px rgba(47,230,200,.4)}
|
| 80 |
+
|
| 81 |
+
/* ---------- scoreboard ---------- */
|
| 82 |
+
.board{flex:1; display:flex; flex-direction:column}
|
| 83 |
+
.cols{display:grid; grid-template-columns:330px 250px 1fr 1fr 1fr; gap:20px; align-items:end;
|
| 84 |
+
padding:0 22px 12px; border-bottom:1px solid var(--line2); margin-bottom:6px}
|
| 85 |
+
.cols span{font-family:var(--mono); font-size:13.5px; font-weight:600; letter-spacing:.2em; color:var(--muted); text-transform:uppercase}
|
| 86 |
+
.cols .num-h{text-align:right}
|
| 87 |
+
.cols i{font-style:normal; color:#7387ab; letter-spacing:.08em; display:block; font-size:11px; margin-top:4px; text-transform:none}
|
| 88 |
+
|
| 89 |
+
.rows{flex:1; display:flex; flex-direction:column; justify-content:space-evenly; gap:8px; padding-top:8px}
|
| 90 |
+
.row{display:grid; grid-template-columns:330px 250px 1fr 1fr 1fr; gap:20px; align-items:center;
|
| 91 |
+
position:relative; padding:14px 22px 14px 26px; border-radius:12px;
|
| 92 |
+
background:linear-gradient(180deg, rgba(16,26,48,.82), rgba(10,16,30,.88));
|
| 93 |
+
border:1px solid var(--line2); box-shadow:0 8px 24px rgba(0,0,0,.12)}
|
| 94 |
+
.row::before{content:""; position:absolute; left:0; top:12px; bottom:12px; width:4px; border-radius:0 3px 3px 0; background:var(--c)}
|
| 95 |
+
|
| 96 |
+
.r-lx {--c:var(--lx)} .r-orn{--c:var(--orn)} .r-q9{--c:var(--q9)} .r-q27{--c:var(--q27)}
|
| 97 |
+
.r-orn{background:linear-gradient(105deg, rgba(199,146,255,.13), rgba(16,26,48,.82) 42%, rgba(10,16,30,.9))}
|
| 98 |
+
|
| 99 |
+
.model .nm{font-family:var(--disp); font-weight:700; text-transform:uppercase; font-size:23px; letter-spacing:.02em; line-height:1; color:var(--c)}
|
| 100 |
+
.model .meta{font-family:var(--mono); font-size:12px; color:var(--dim); margin-top:7px; letter-spacing:.04em}
|
| 101 |
+
|
| 102 |
+
.speed .v{font-family:var(--mono); font-weight:700; font-size:36px; line-height:1; text-align:right; color:var(--text)}
|
| 103 |
+
.speed.win .v{color:var(--c); text-shadow:0 0 24px color-mix(in srgb, var(--c) 50%, transparent)}
|
| 104 |
+
.speed.win .v::before{content:"▲"; font-size:13px; margin-right:9px; vertical-align:4px}
|
| 105 |
+
.speed .pp{font-family:var(--mono); font-weight:700; font-size:17px; text-align:right; color:var(--ls);
|
| 106 |
+
margin-top:5px; letter-spacing:.01em; text-shadow:0 0 16px rgba(47,230,200,.35)}
|
| 107 |
+
.speed .pp small{font-size:10px; font-weight:600; color:var(--dim); letter-spacing:.14em; text-transform:uppercase; margin-left:5px}
|
| 108 |
+
.speed .bar{height:8px; margin-top:9px; background:rgba(9,14,26,.85); border:1px solid var(--line2); border-radius:4px; overflow:hidden}
|
| 109 |
+
.speed .bar i{display:block; height:100%; width:var(--w); border-radius:3px;
|
| 110 |
+
background:linear-gradient(90deg, color-mix(in srgb, var(--c) 25%, transparent), var(--c));
|
| 111 |
+
box-shadow:0 0 12px color-mix(in srgb, var(--c) 60%, transparent)}
|
| 112 |
+
|
| 113 |
+
.cell{text-align:right}
|
| 114 |
+
.cell .v{font-family:var(--mono); font-weight:600; font-size:32px; line-height:1; color:var(--muted)}
|
| 115 |
+
.cell .v .st{font-size:16px; color:var(--dim); vertical-align:6px; margin-left:2px}
|
| 116 |
+
.cell.na .v{color:#3d4c66}
|
| 117 |
+
.cell.win .v{color:var(--c); font-weight:700; text-shadow:0 0 22px color-mix(in srgb, var(--c) 45%, transparent)}
|
| 118 |
+
.cell.win .v::before{content:"▲"; font-size:13px; margin-right:9px; vertical-align:4px}
|
| 119 |
+
.cell.win{background:color-mix(in srgb, var(--c) 9%, transparent);
|
| 120 |
+
border:1px solid color-mix(in srgb, var(--c) 30%, transparent);
|
| 121 |
+
border-radius:9px; padding:12px 16px; margin:-12px -16px}
|
| 122 |
+
|
| 123 |
+
/* ---------- verdict strip ---------- */
|
| 124 |
+
.verdict{display:flex; align-items:stretch; margin-top:14px; border:1px solid var(--line); border-radius:12px;
|
| 125 |
+
overflow:hidden; background:linear-gradient(180deg, rgba(16,26,48,.92), rgba(10,16,30,.94));
|
| 126 |
+
box-shadow:0 10px 30px rgba(0,0,0,.2)}
|
| 127 |
+
.vseg{display:flex; flex-direction:column; justify-content:center; gap:4px; padding:11px 24px; position:relative;
|
| 128 |
+
transition:background .25s}
|
| 129 |
+
.vseg:hover{background:rgba(140,170,220,.05)}
|
| 130 |
+
.vseg + .vseg{border-left:1px solid var(--line2)}
|
| 131 |
+
.v-sweep{flex:1.7; background:linear-gradient(105deg, rgba(77,184,255,.10), transparent 55%)}
|
| 132 |
+
.v-range{flex:1.25}
|
| 133 |
+
.v-speed{flex:1}
|
| 134 |
+
.vk{font-family:var(--disp); font-weight:700; text-transform:uppercase; font-size:15px; letter-spacing:.06em;
|
| 135 |
+
display:flex; align-items:center; gap:9px}
|
| 136 |
+
.vk::before{content:""; width:5px; height:15px; border-radius:2px; background:currentColor; box-shadow:0 0 10px currentColor}
|
| 137 |
+
.v-sweep .vk{color:var(--lx)} .v-range .vk{color:var(--orn)} .v-speed .vk{color:var(--ls)}
|
| 138 |
+
.vd{font-family:var(--mono); font-size:12.5px; color:var(--muted); letter-spacing:.02em; padding-left:14px}
|
| 139 |
+
.vd b{font-weight:700}
|
| 140 |
+
.v-sweep .vd b{color:var(--lx)} .v-range .vd b{color:var(--orn)} .v-speed .vd b{color:var(--ls)}
|
| 141 |
+
.vd .up{color:var(--ls)} .vd .dn{color:#7d8ba3}
|
| 142 |
+
|
| 143 |
+
/* ---------- footer ---------- */
|
| 144 |
+
footer{display:flex; justify-content:space-between; align-items:flex-end; gap:24px; margin-top:16px;
|
| 145 |
+
font-family:var(--mono); font-size:12px; color:var(--dim); letter-spacing:.03em}
|
| 146 |
+
footer .cav b{color:var(--muted); font-weight:500}
|
| 147 |
+
footer .cav .g{color:var(--ls)}
|
| 148 |
+
footer .cav .tri{color:var(--q27)}
|
| 149 |
+
footer .brand b{color:var(--muted); font-weight:600}
|
| 150 |
+
</style>
|
| 151 |
+
</head>
|
| 152 |
+
<body>
|
| 153 |
+
<div class="stage">
|
| 154 |
+
|
| 155 |
+
<header>
|
| 156 |
+
<div>
|
| 157 |
+
<div class="eyebrow">TurboQuant · TQ3_4S · single RTX 3090</div>
|
| 158 |
+
<h1>Laguna XS 2.1-<span class="acc">TQ3_4S</span><br><span class="acc2">3B active. Fastest here.</span></h1>
|
| 159 |
+
<div class="headstats">
|
| 160 |
+
<div class="hstat hx"><b>100%</b><span>coding suite</span></div>
|
| 161 |
+
<div class="hstat hs"><b>96.4%</b><span>speed suite</span></div>
|
| 162 |
+
<div class="hstat"><b>3B</b><span>active · 30B MoE</span></div>
|
| 163 |
+
<div class="hstat"><b>4.05</b><span>bits / weight</span></div>
|
| 164 |
+
</div>
|
| 165 |
+
</div>
|
| 166 |
+
<div class="marquee">
|
| 167 |
+
<div class="big">196<small>tok/s</small></div>
|
| 168 |
+
<div class="mlabel"><b>decode</b> · greedy</div>
|
| 169 |
+
<div class="big2">1389<small>tok/s</small></div>
|
| 170 |
+
<div class="mlabel2"><b>pp128</b> · prefill</div>
|
| 171 |
+
</div>
|
| 172 |
+
</header>
|
| 173 |
+
|
| 174 |
+
<div class="board">
|
| 175 |
+
<div class="cols">
|
| 176 |
+
<span>Model</span>
|
| 177 |
+
<span class="num-h">Decode<i>tok/s · greedy</i></span>
|
| 178 |
+
<span class="num-h">HumanEval+<i>pass@1</i></span>
|
| 179 |
+
<span class="num-h">MBPP+<i>pass@1</i></span>
|
| 180 |
+
<span class="num-h">Hard86<i>86 assertions</i></span>
|
| 181 |
+
</div>
|
| 182 |
+
|
| 183 |
+
<div class="rows">
|
| 184 |
+
<div class="row r-lx">
|
| 185 |
+
<div class="model"><div class="nm">Laguna XS 2.1</div><div class="meta">TQ3_4S · 4.05 bpw · RTX 3090</div></div>
|
| 186 |
+
<div class="speed win"><div class="v">196</div><div class="pp">1389<small>pp128</small></div><div class="bar"><i style="--w:100%"></i></div></div>
|
| 187 |
+
<div class="cell"><span class="v">76.2<span class="st">%</span></span></div>
|
| 188 |
+
<div class="cell"><span class="v">72.0<span class="st">%</span></span></div>
|
| 189 |
+
<div class="cell"><span class="v">64.0<span class="st">%</span></span></div>
|
| 190 |
+
</div>
|
| 191 |
+
|
| 192 |
+
<div class="row r-orn">
|
| 193 |
+
<div class="model"><div class="nm">Ornith 35B</div><div class="meta">TQ3_4S · 13.0 GiB · RTX 3090</div></div>
|
| 194 |
+
<div class="speed"><div class="v">146.3</div><div class="bar"><i style="--w:74.6%"></i></div></div>
|
| 195 |
+
<div class="cell"><span class="v">87.2<span class="st">%</span></span></div>
|
| 196 |
+
<div class="cell"><span class="v">70.9<span class="st">%</span></span></div>
|
| 197 |
+
<div class="cell"><span class="v">81.4<span class="st">%</span></span></div>
|
| 198 |
+
</div>
|
| 199 |
+
|
| 200 |
+
<div class="row r-q9">
|
| 201 |
+
<div class="model"><div class="nm">Qwen3.5 9B</div><div class="meta">TQ3_4S · RTX 3090</div></div>
|
| 202 |
+
<div class="speed"><div class="v">134</div><div class="bar"><i style="--w:68.4%"></i></div></div>
|
| 203 |
+
<div class="cell"><span class="v">67.1<span class="st">%</span></span></div>
|
| 204 |
+
<div class="cell"><span class="v">56.3<span class="st">%</span></span></div>
|
| 205 |
+
<div class="cell"><span class="v">44.2<span class="st">%</span></span></div>
|
| 206 |
+
</div>
|
| 207 |
+
|
| 208 |
+
<div class="row r-q27">
|
| 209 |
+
<div class="model"><div class="nm">Qwen3.6 27B MTP</div><div class="meta">TQ3_4S · RTX 3090</div></div>
|
| 210 |
+
<div class="speed"><div class="v">54</div><div class="bar"><i style="--w:27.6%"></i></div></div>
|
| 211 |
+
<div class="cell win"><span class="v">92.7<span class="st">%*</span></span></div>
|
| 212 |
+
<div class="cell win"><span class="v">87.8<span class="st">%*</span></span></div>
|
| 213 |
+
<div class="cell win"><span class="v">88.4<span class="st">%</span></span></div>
|
| 214 |
+
</div>
|
| 215 |
+
</div>
|
| 216 |
+
</div>
|
| 217 |
+
|
| 218 |
+
<footer>
|
| 219 |
+
<div class="cav"><span class="tri">▲</span> best in column · EvalPlus = greedy pass@1 · <b>*</b> 27B via <span class="g">vLLM-nvfp4</span> (directional) · all rows on one RTX 3090</div>
|
| 220 |
+
<div class="brand"><b>turbo-tan/llama.cpp-tq3</b> · greedy · FA on · 2026-07-31</div>
|
| 221 |
+
</footer>
|
| 222 |
+
|
| 223 |
+
</div>
|
| 224 |
+
</body>
|
| 225 |
+
</html>
|