How do I install the Llmbench skill?

Run npx skillmds@latest add az9713/llmbench in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Llmbench safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Llmbench?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Llmbench free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Llmbench?

az9713 (@az9713) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Llmbench by az9713 — Agent Skill

Name: Llmbench
Author: az9713

File contents

LLM Benchmark Pipeline

Full pipeline: parallel model testing + Remotion 2x2 grid video comparison.

Default Models

openrouter/z-ai/glm-5 -> llmtest/game-glm5.html
openrouter/minimax/minimax-m2.5 -> llmtest/game-minimax25.html
openrouter/google/gemini-3-pro-preview -> llmtest/game-gemini3pro.html
openrouter/anthropic/claude-opus-4.6 -> llmtest/game-opus46.html

Pipeline Stages

Stage 1: Parallel Model Execution

Source environment: set -a && source .env 2>/dev/null; set +a
Clean previous outputs in llmtest/
Run all 4 models in parallel as background processes:

opencode run -m openrouter/z-ai/glm-5 "$PROMPT Save the complete file as llmtest/game-glm5.html" > /tmp/opencode-glm5.log 2>&1 &
PID1=$!
opencode run -m openrouter/minimax/minimax-m2.5 "$PROMPT Save the complete file as llmtest/game-minimax25.html" > /tmp/opencode-minimax25.log 2>&1 &
PID2=$!
opencode run -m openrouter/google/gemini-3-pro-preview "$PROMPT Save the complete file as llmtest/game-gemini3pro.html" > /tmp/opencode-gemini3pro.log 2>&1 &
PID3=$!
opencode run -m openrouter/anthropic/claude-opus-4.6 "$PROMPT Save the complete file as llmtest/game-opus46.html" > /tmp/opencode-opus46.log 2>&1 &
PID4=$!
wait $PID1 $PID2 $PID3 $PID4

Verify all 4 HTML files exist and are >1KB

Stage 2: Copy HTML to Remotion

cp llmtest/game-*.html remotion/public/

Stage 3: Render Video

cd remotion && npx remotion render index.ts GridComparison ../out/comparison.mp4 --concurrency=1 --timeout=90000 --public-dir=public

--concurrency=1 is required because HTML files use requestAnimationFrame canvas animations
Remotion can't sync with those, so sequential single-tab rendering lets animations advance naturally

Stage 4: Verify Output

Check out/comparison.mp4 exists
Report file size
Report success/failure

Important Notes

Remotion must be installed first: cd remotion && npm install
The HTML files must use only vanilla HTML/CSS/JS (no external dependencies)
Video output: 1920x1080, 30fps, 30 seconds (900 frames)
Each quadrant shows one model's output with a label bar at the bottom

---name: llmbenchdescription: Full LLM benchmark pipeline — runs same prompt across 4 models in parallel via OpenCode, then renders a 2x2 grid comparison video using Remotion.---# LLM Benchmark PipelineFull pipeline: parallel model testing + Remotion 2x2 grid video comparison.## Default Models- `openrouter/z-ai/glm-5` -> `llmtest/game-glm5.html`- `openrouter/minimax/minimax-m2.5` -> `llmtest/game-minimax25.html`- `openrouter/google/gemini-3-pro-preview` -> `llmtest/game-gemini3pro.html`- `openrouter/anthropic/claude-opus-4.6` -> `llmtest/game-opus46.html`## Pipeline Stages### Stage 1: Parallel Model Execution1. Source environment: `set -a && source .env 2>/dev/null; set +a`2. Clean previous outputs in `llmtest/`3. Run all 4 models in parallel as background processes:```bashopencode run -m openrouter/z-ai/glm-5 "$PROMPT Save the complete file as llmtest/game-glm5.html" > /tmp/opencode-glm5.log 2>&1 &PID1=$!opencode run -m openrouter/minimax/minimax-m2.5 "$PROMPT Save the complete file as llmtest/game-minimax25.html" > /tmp/opencode-minimax25.log 2>&1 &PID2=$!opencode run -m openrouter/google/gemini-3-pro-preview "$PROMPT Save the complete file as llmtest/game-gemini3pro.html" > /tmp/opencode-gemini3pro.log 2>&1 &PID3=$!opencode run -m openrouter/anthropic/claude-opus-4.6 "$PROMPT Save the complete file as llmtest/game-opus46.html" > /tmp/opencode-opus46.log 2>&1 &PID4=$!wait $PID1 $PID2 $PID3 $PID4```4. Verify all 4 HTML files exist and are >1KB### Stage 2: Copy HTML to Remotion```bashcp llmtest/game-*.html remotion/public/```### Stage 3: Render Video```bashcd remotion && npx remotion render index.ts GridComparison ../out/comparison.mp4 --concurrency=1 --timeout=90000 --public-dir=public```- `--concurrency=1` is required because HTML files use requestAnimationFrame canvas animations- Remotion can't sync with those, so sequential single-tab rendering lets animations advance naturally### Stage 4: Verify Output1. Check `out/comparison.mp4` exists2. Report file size3. Report success/failure## Important Notes- Remotion must be installed first: `cd remotion && npm install`- The HTML files must use only vanilla HTML/CSS/JS (no external dependencies)- Video output: 1920x1080, 30fps, 30 seconds (900 frames)- Each quadrant shows one model's output with a label bar at the bottom

Frequently asked questions

Llmbench