CLI Commands

The modellens CLI is the primary entry point for Model Lens. It’s built with Click and delegates to the appropriate benchmark framework.

Global options

modellens --help
modellens --version

run — Run benchmarks

modellens run [OPTIONS]

Option	Type	Default	Description
--provider, -p	choice	auto-detect	lm-studio, ollama, open-webui, jan, llama.cpp, vllm
--models, -m	multiple	auto-detect	Models to benchmark (repeatable)
--model-name	string	—	Single model name (alias for --models)
--framework, -f	choice	devbench	devbench, general, compare
--output-dir, -o	string	results	Output directory
--num-runs, -n	int	5	Runs per prompt (devbench)
--num-prompts	int	20	Prompt count
--samples	int	—	Samples per benchmark (general)
--quick	flag	—	Quick mode: fewer prompts and runs
--config, -c	string	—	Config file path
--parallel/--sequential	flag	--parallel	Parallel execution
--seed	int	—	Random seed
--verbose, -v	flag	—	Show raw model responses
--ci	flag	—	CI-safe headless mode
--traces-dir	string	results/traces	Trace capture directory
--no-traces	flag	—	Disable trace capture
--sse-port	int	—	Enable SSE streaming on port N
--json-output	flag	—	Output results as JSON to stdout

Examples

# Auto-detect provider and models, quick mode
modellens run --quick

# Ollama with specific models
modellens run --provider ollama --models llama3.2

# General benchmark framework
modellens run --framework general --model-name my-model --samples 50

# Model comparison (both frameworks)
modellens run --framework compare --models qwen3.5 gemma-4

# SSE streaming for dashboard
modellens run --sse-port 9090

# CI mode with JSON output
modellens run --ci --json-output > results.json

---

workload — Workload evaluation

modellens workload run [OPTIONS]

Option	Type	Default	Description
--model, -m	string	(required)	Model to evaluate
--project, -p	string	auto-detect	Project path to evaluate
--output-dir, -o	string	results/workload	Output directory

Example

modellens workload run --model qwen3.5-9b
modellens workload run --model qwen3.5-9b --project /path/to/react-app

---

info — System information

modellens info

Displays:

• Hardware details (CPU, GPU, RAM, OS, architecture)
• Python version
• Connected providers and their status
• Available models on each provider

---

models — List available models

modellens models [OPTIONS]

Option	Type	Default	Description
--provider, -p	choice	auto-detect	Filter by provider

Example

modellens models
modellens models --provider ollama

---

leaderboard — View benchmark results

modellens leaderboard [OPTIONS] [RESULTS_DIR]

Option	Type	Default	Description
--format	choice	table	table, json, csv
--sort-by	string	overall_score	Sort column
--limit	int	20	Max results

Example

modellens leaderboard results/
modellens leaderboard --format json --sort-by tokens_per_sec

---

health — Provider health check

modellens health [OPTIONS]

Option	Type	Default	Description
--provider, -p	choice	auto-detect	Provider to check

Example

modellens health
modellens health --provider ollama

---

compare — Model comparison

modellens compare [OPTIONS] MODEL_A MODEL_B

Runs both models against the same benchmarks and produces side-by-side results.

Example

modellens compare qwen3.5-9b gemma-4-9b