perplexity-cli-models

Model selection guide for perplexity-cli ask and chat commands. Use when choosing between sonar, sonar-pro, sonar-reasoning, and sonar-deep-research.

Quality

90%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

perplexity-cli — Model Selection

The --model flag (or "model" in JSON input) controls which Sonar model handles your question. Applies to ask and chat. Default is sonar.

Model Comparison

Model	Speed	Cost	Best for
`sonar`	Fastest	Lowest	Quick factual questions, high-volume agent queries
`sonar-pro`	Medium	Medium	Deep analysis, comparisons, multi-part questions
`sonar-reasoning`	Slow	High	Complex analytical or logical questions
`sonar-deep-research`	Slowest	Highest	Extensive research with multiple search rounds

When to Use Each

`sonar` (default)

Use for:

Simple factual lookups: "What is the current version of Node.js?"
High-volume batch queries where cost matters
Quick context-gathering before a more complex follow-up
Any question where speed > depth

perplexity-cli ask "What is the latest LTS version of Node.js?"
perplexity-cli ask "What does HTTP 429 mean?"
perplexity-cli search "Python 3.13 release date"

`sonar-pro`

Use for:

Comparative analysis: "Compare X vs Y"
Multi-step reasoning where you want more than a surface answer
Technical architecture questions
Summaries of complex topics with citations
When you want higher quality without the latency of reasoning models

perplexity-cli ask "Compare PostgreSQL vs CockroachDB for a multi-region SaaS" --model sonar-pro
perplexity-cli ask "What are the tradeoffs of event sourcing vs CQRS?" --model sonar-pro
perplexity-cli ask "Summarize recent changes in Kubernetes 1.30" --model sonar-pro --recency month

`sonar-reasoning`

Use for:

Questions that require logical deduction
Algorithm complexity analysis
Debugging assistance for tricky problems
Math or formal reasoning
Combine with --reasoning high for best results

perplexity-cli ask "What is the time complexity of Dijkstra's algorithm and why?" \
  --model sonar-reasoning --reasoning high

perplexity-cli ask "Given a distributed system with eventual consistency, explain why this race condition occurs: ..." \
  --model sonar-reasoning --reasoning high

`sonar-deep-research`

Use for:

Comprehensive literature reviews
Research reports that need to synthesize many sources
Questions where coverage > speed
Generating reference material rather than quick answers
Long wait times are acceptable

perplexity-cli ask "What does the current research literature say about transformer attention mechanisms and their computational limits?" \
  --model sonar-deep-research --mode academic

perplexity-cli ask "Provide a comprehensive overview of WebAssembly ecosystem tools as of 2025" \
  --model sonar-deep-research

Reasoning Effort (`--reasoning`)

Only meaningful with sonar-reasoning and sonar-pro. Ignored by other models.

Value	Use when
`minimal`	Speed is critical; shallow reasoning is fine
`low`	Slightly better than default; marginal cost increase
`medium`	Balanced — good default for `sonar-reasoning`
`high`	Complex questions where correctness matters more than speed

# Quick estimation
perplexity-cli ask "Rough complexity of merge sort?" --model sonar-reasoning --reasoning minimal

# Deep analysis
perplexity-cli ask "Analyze the correctness of this distributed locking algorithm: ..." \
  --model sonar-reasoning --reasoning high

Decision Flowchart

Is it a simple factual question?
  → YES: sonar (fast, cheap)
  → NO:
    Does it need logical/mathematical reasoning?
      → YES: sonar-reasoning + --reasoning high
      → NO:
        Does it need comprehensive coverage from many sources?
          → YES: sonar-deep-research
          → NO: sonar-pro (deep analysis, balanced speed/cost)

Cost-Aware Agent Usage

For agents making many queries in a loop:

# Use sonar for bulk/filter queries
for topic in "Python" "Rust" "Go" "TypeScript"; do
  perplexity-cli ask "Latest stable version of $topic?" | jq -r '.content'
done

# Upgrade to sonar-pro only for the synthesis step
perplexity-cli ask "Given these versions, which language has the most active recent development?" \
  --model sonar-pro \
  --system "Be brief. One paragraph max."

Repository: slurpyb/perplexity-cli
Path: .claude/skills/web-search-with-perplexity/models/SKILL.md
Commit: 5624eaa

Last updated: 2 days ago
First committed: 4 months ago

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.