AI Rankings Leaderboard — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited AI Rankings Leaderboard (Agent Skill) and scored it 92/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 2 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 2 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
A comprehensive skill for querying AI model and application rankings from multiple authoritative sources. Get the latest insights on LLM performance, popularity, pricing, and value metrics.
| Source | URL | Focus |
|---|---|---|
| Artificial Analysis | https://artificialanalysis.ai/ | Intelligence Index, Speed, Price benchmarks |
| LLM Leaderboard | https://artificialanalysis.ai/leaderboards/models | Model comparison (100+ models) |
| LLM API Providers | https://artificialanalysis.ai/leaderboards/providers | API Provider comparison (500+ endpoints) |
| Image & Video Leaderboards | https://artificialanalysis.ai/ (Image & Video section) | Image/Video model ELO rankings |
| OpenRouter Rankings | https://openrouter.ai/rankings | Model usage & popularity |
| OpenRouter Apps | https://openrouter.ai/apps | AI applications ranking |
| OpenRouter Models | https://openrouter.ai/models | All available models with pricing |
| OpenRouter Free Models | https://openrouter.ai/models?q=free | Free models only |
| Pinchbench | https://pinchbench.com/ | Model benchmark (Success Rate, Speed, Cost, Value) |
Intelligence Index (智力指数)
Artificial Analysis Coding Index (编程能力指数)
Artificial Analysis Agentic Index (智能体能力指数)
Performance Metrics
| Metric | Description |
|---|---|
| Intelligence Index | Overall model intelligence score (higher is better) |
| Speed | Output tokens per second (tokens/s) |
| Blended Price | Combined USD per million tokens (3:1 input/output ratio) |
| Input Price | Price per million input tokens (USD) |
| Output Price | Price per million output tokens (USD) |
| Latency (TTFT) | Time to First Token in seconds |
| Context Window | Maximum context length supported |
Model Comparison Table Columns
| Column | Description |
|---|---|
| Features | Model features (reasoning badge, etc.) |
| Model | Model name with logo |
| Context Window | Max context length |
| Creator | Provider/Company |
| Intelligence Index | AI intelligence score |
| Blended USD/1M Tokens | Combined input/output price |
| Median Tokens/s | Median output speed |
| Latency First Chunk (s) | Time to first token |
| Further Analysis | Link to detailed analysis |
Filters Available
| Filter | Options |
|---|---|
| Frontier Models | On/Off |
| Open Weights | On/Off (开源权重模型) |
| Size Class | Small, Medium, Large, etc. |
| Reasoning | On/Off (推理模型筛选) |
| Model Status | Current, Preview, Discontinued |
Comparison of 500+ AI Model Endpoints
| Column | Description |
|---|---|
| API Provider | Provider name (Cerebras, Groq, Fireworks, etc.) |
| Model | Model name |
| Context Window | Max context length |
| License | Model license |
| Intelligence Index | Model intelligence score |
| Blended USD/1M Tokens | Combined price |
| Median Tokens/s | Output speed |
| Median First Chunk (s) | Latency (TTFT) |
| Total Response (s) | End-to-end response time |
| Reasoning Time (s) | Reasoning model computation time |
| End-to-End Response Time | Full request-response cycle |
Key Providers
Text-to-Image Leaderboard
Video Leaderboards
| Category | Description |
|---|---|
| Text to Video (with Audio) | Text generates video with sound |
| Text to Video (without Audio) | Text generates silent video |
| Image to Video (with Audio) | Image + text generates video with sound |
| Image to Video (without Audio) | Image + text generates silent video |
| Image Editing | Edit existing images with AI |
Evaluation Method
model parameter to use when calling OpenRouter APIThis skill requires:
execute_command: Execute shell commands and scriptsuse_skill: Load browser-automation skill for JavaScript-rendered pagesweb_fetch: Fallback for simple HTTP requestsRequired CLI Dependency: agent-browser
The agent-browser CLI must be installed before using this skill. Install via:
npm install -g agent-browser
# or
npx agent-browser --versionThis skill calls agent-browser via subprocess with hardcoded argument arrays (no shell injection risk).
Note on browser eval: The agent-browser eval command executes document.body.innerText or similar DOM queries on the remote page to extract rendered content. This is standard web scraping behavior for JavaScript-rendered pages and is limited to reading page content only.
For JavaScript-rendered pages (OpenRouter Rankings, Artificial Analysis), this skill uses browser automation:
use_skill("browser-automation") agent-browser open "https://artificialanalysis.ai/leaderboards/models"
agent-browser wait --load networkidle
agent-browser eval "document.body.innerText"https://artificialanalysis.ai/leaderboards/models - LLM comparison (100+ models)https://artificialanalysis.ai/leaderboards/providers - API providers (500+ endpoints)https://artificialanalysis.ai/ - Image & Video leaderboardshttps://openrouter.ai/rankings - Model usage rankings (JS rendered)https://openrouter.ai/apps - App rankings (JS rendered)LLM Leaderboard Page (/leaderboards/models):
LLM Leaderboard - Comparison of over 100 AI models
├── HIGHLIGHTS section
│ ├── Intelligence: Gemini 3.1 Pro Preview, GPT-5.4 (xhigh)
│ ├── Speed: Mercury 2 (943 t/s), NVIDIA Nemotron 3 Super (462 t/s)
│ └── Price: Gemma 3n E4B (cheapest)
├── Filters:
│ ├── Frontier Models | Open Weights | Size Class | Reasoning | Model Status
├── Comparison table columns:
│ ├── Features | Model | Context Window | Creator
│ ├── Intelligence Index | Blended USD/1M | Median Tokens/s | Latency
│ └── Further Analysis
└── Key definitions (expandable)
├── Context window
├── Output Speed (tokens/s)
├── Latency (Time to First Token)
├── Price (3:1 blended)
├── Output Price
└── Input PriceLLM API Providers Page (/leaderboards/providers):
LLM API Providers Leaderboard - 500+ endpoints
├── Filters (same as LLM Leaderboard)
├── Comparison table columns:
│ ├── API Provider | Model | Context Window | License
│ ├── Intelligence Index | Blended USD/1M | Median Tokens/s
│ ├── Median First Chunk (s) | Total Response (s) | Reasoning Time (s)
│ └── Further Analysis
└── 24+ Providers: Cerebras, Groq, Fireworks, SambaNova, etc.Image & Video Leaderboards (on homepage):
Image & Video Leaderboards
├── Tabs:
│ ├── Text to Image (ELO scores, 95% CI)
│ ├── Image Editing
│ ├── Text to Video (with Audio)
│ ├── Text to Video (without Audio)
│ ├── Image to Video (with Audio)
│ └── Image to Video (without Audio)
└── Top models with ELO rankingsOpenRouter Rankings Page (/rankings):
https://openrouter.ai/rankings
├── Top Models (chart header)
├── LLM Leaderboard ← THIS is the usage ranking (parse this!)
│ ├── 1. MiniMax M2.5 (1.75T tokens)
│ ├── 2. Step 3.5 Flash (1.34T tokens)
│ └── [Show more] button
├── Market Share (different metric - don't mix!)
└── ...User: "What are the top models on Artificial Analysis Intelligence Index?"
-> Fetches Artificial Analysis LLM Leaderboard and displays top models by intelligenceUser: "Which AI models are the fastest in terms of output speed?"
-> Fetches Artificial Analysis data and lists models by tokens/secondUser: "Compare LLM API providers like Cerebras and Groq"
-> Fetches Artificial Analysis Providers Leaderboard and compares speed/priceUser: "What are the best text-to-image models?"
-> Fetches Artificial Analysis Image Arena leaderboard with ELO scoresUser: "What are the top 10 AI models right now?"
-> Fetches OpenRouter rankings and displays top models with usage statsUser: "What free models are available on OpenRouter?"
-> Fetches https://openrouter.ai/models?q=free and lists all free models with their model IDsUser: "What's the model ID for GPT-4o on OpenRouter?"
-> Fetches https://openrouter.ai/models and returns the exact model parameter to useUser: "Compare GPT-4 and Claude on Pinchbench"
-> Fetches Pinchbench data and compares success rate, speed, cost==================================================
Artificial Analysis Intelligence Index
==================================================
Top 10 Models by Intelligence:
| Rank | Model | Intelligence | Speed (t/s) | Price ($/M) |
|------|-------|--------------|-------------|-------------|
| 1 | Gemini 3.1 Pro Preview | 57 | ~50 | $1.25 |
| 2 | GPT-5.4 (xhigh) | 57 | ~60 | $15.00 |
| 3 | Claude Opus 4.6 (max) | 53 | ~80 | $18.00 |
| 4 | Claude Sonnet 4.6 (max) | 52 | ~85 | $4.50 |
| 5 | GLM-5 | 50 | ~45 | $0.50 |
...
Fastest Models: Mercury 2 (943 t/s), NVIDIA Nemotron 3 Super (462 t/s)
Best Price: Gemma 3n E4B, Granite 4.0 H Small
Data Source: Artificial Analysis (artificialanalysis.ai)
====================================================================================================
LLM API Providers Leaderboard
==================================================
| Provider | Model | Speed (t/s) | Price ($/M) | Latency (s) |
|----------|-------|-------------|-------------|-------------|
| Cerebras | Llama 3.1 70B | 2143 | $0.12 | 0.08 |
| Groq | Llama 3.1 70B | 943 | $0.59 | 0.15 |
| Fireworks | Llama 3.1 70B | 562 | $0.90 | 0.22 |
...
Data Source: Artificial Analysis Providers
====================================================================================================
Text-to-Image Leaderboard (ELO)
==================================================
| Rank | Model | ELO Score | 95% CI |
|------|-------|-----------|--------|
| 1 | GPT Image 1.5 (high) | 1342 | ±12 |
| 2 | Imagen 4 Ultra | 1289 | ±15 |
| 3 | Gemini 3.1 Flash Image | 1245 | ±18 |
...
Data Source: Artificial Analysis Image Arena
====================================================================================================
AI Model Rankings (OpenRouter)
==================================================
Top 10 Models by Usage:
| Rank | Model | Provider | Tokens | Growth |
|------|-------|----------|--------|--------|
| 1 | MiniMax M2.5 | minimax | 1.75T | +15% |
| 2 | Step 3.5 Flash | step | 1.34T | +22% |
...
Data Source: OpenRouter (Weekly Rankings)
====================================================================================================
Free Models on OpenRouter
==================================================
| Model Name | Model ID (for API) | Context |
|------------|-------------------|---------|
| GPT-4o Mini | openai/gpt-4o-mini | 128K |
| Llama 3.3 70B | meta-llama/llama-3.3-70b-instruct | 128K |
| DeepSeek V3 | deepseek/deepseek-chat | 64K |
...
💡 Usage: Set model parameter to the Model ID value
Example: model="openai/gpt-4o-mini"
Data Source: OpenRouter Models
==================================================Artificial Analysis and OpenRouter rankings pages require JavaScript rendering:
# Step 1: Load browser-automation skill (REQUIRED)
use_skill("browser-automation")
# Step 2: Navigate to Artificial Analysis LLM Leaderboard
agent-browser open "https://artificialanalysis.ai/leaderboards/models"
agent-browser wait --load networkidle
# Step 3: Wait for content to load, then extract
agent-browser wait 3000
agent-browser eval "document.body.innerText"
# Step 4: Close browser when done
agent-browser closeUse the query_leaderboard.py script to fetch model data via OpenRouter API (no JavaScript needed):
# List free models
python3 "${SKILL_DIR}/query_leaderboard.py --free"
# Search models by name
python3 "${SKILL_DIR}/query_leaderboard.py -s glm"
python3 "${SKILL_DIR}/query_leaderboard.py -s gpt"
# Get specific model info
python3 "${SKILL_DIR}/query_leaderboard.py --id openai/gpt-4o"
# List all models with limit
python3 "${SKILL_DIR}/query_leaderboard.py --all --limit 50"When browser/Python is not available, use web_fetch:
https://artificialanalysis.ai/leaderboards/modelshttps://openrouter.ai/api/v1/modelshttps://pinchbench.com/Note: Rankings pages require JavaScript rendering - use browser automation (Method 1).
openai/gpt-4o-mini) as the model parameter in API callsBased on observed page structure, Artificial Analysis provides:
/models/{model-id}/providers endpoint patternExample model providers API:
/models/gpt-oss-120b/providers
/models/gemini-3-1-pro-preview/providers
/models/claude-opus-4-6-adaptive/providersWhen calling OpenRouter API (for chat completions), use the Model ID. Note: This skill's scripts (fetch_rankings.py, query_leaderboard.py) only read public leaderboard data and do NOT require API authentication.
curl https://openrouter.ai/api/v1/chat/completions \
-H "Authorization: Bearer $OPENROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "openai/gpt-4o-mini", # <- Model ID from this skill
"messages": [{"role": "user", "content": "Hello"}]
}'~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.