Home / Model Guide
Model Guide

Which AI for which job.

Not all AI tools are equal. Picking the wrong one wastes money and produces bad output. This is the Talkerstein model selection guide — built from real client work, not benchmark marketing.

Model Selection by Use Case
Deep strategy & complex reasoning

ChatGPT GPT-5.5 Thinking / Claude Opus 4.7

Strong for difficult professional work, reasoning, long-horizon analysis, and agentic tasks.

BrevityMedium or high reasoning. Use xhigh only for highest-value complex work.
Best ForStrategy, audits, business plans, technical diagnosis, agentic workflows.
Avoid WhenSimple captions, commodity summaries, or low-risk tasks where speed matters more.
Everyday marketing & client deliverables

ChatGPT GPT-5.5 / Claude Sonnet 4.6

Balanced quality for copy, briefs, strategy notes, client-facing docs, and iterative edits.

BrevityStandard or Detailed. Ask for examples, constraints, and a final QA pass.
Best ForCampaigns, social posts, emails, landing page sections, service pages.
Avoid WhenFresh research claims without browsing or citations.
Fast drafting & high-volume lightweight work

Claude Haiku 4.5 / Gemini Flash or Flash-Lite

Useful when speed and volume matter more than deep reasoning.

BrevityConcise. Strong templates. Batch tasks.
Best ForClassification, tagging, rewriting, simple extraction, early drafts.
Avoid WhenHigh-stakes strategy, complex code, legal/compliance-heavy work.
Current research, competitors, trends, facts

Perplexity Sonar / Agent API

Designed for web-grounded answers, real-time retrieval, and citations.

BrevityDetailed with source list. Require exact dates and URLs.
Best ForMarket research, competitor pages, pricing, policy checks, current events.
Avoid WhenTasks that use only private files or internal data.
SEO / AEO / GEO content production

ChatGPT GPT-5.5 + Perplexity Sonar + Claude Sonnet

Combines reasoning, current source-gathering, and clean long-form writing.

BrevityDetailed. Force search intent, internal links, FAQs, schema, and citations.
Best ForArticles, topical maps, service pages, answer-engine-ready copy.
Avoid WhenPublishing current claims without source verification.
Image prompt engineering & editable design

Gemini Image Models / Canva AI / ChatGPT for prompt refinement

Use a text model for prompt strategy and image/design tools for production and editing.

BrevityExact Spec. Include aspect ratio, camera, style, lighting, negative prompt.
Best ForAds, social graphics, brand mockups, product visuals.
Avoid WhenWhen exact brand/legal usage rights are not confirmed.
Video prompt engineering & AI UGC ads

Higgsfield for generation + ChatGPT/Claude for storyboarding

Video outputs need scene logic, motion progression, platform-native framing, and negative prompts.

BrevityExact Spec. 8-second modules work well for social testing.
Best ForUGC-style ads, product moments, hero video concepts, creative tests.
Avoid WhenComplex factual claims or brand claims that should be scripted separately.
Spreadsheet, data analysis & structured output

ChatGPT GPT-5.5 Thinking / Gemini 3.1 Pro Preview

Good for structured reasoning, tool use, and multi-step analysis when data is supplied.

BrevityDetailed. Require assumptions, formulas, checks, and decision summary.
Best ForDashboards, data interpretation, KPI reviews, campaign analysis.
Avoid WhenDirty datasets without definitions or source context.
Brevity Modes — Set Depth Before You Prompt

The instruction you give at the start of a prompt determines how much output you get. Picking the wrong brevity wastes time on output you'll edit down — or leaves you with too little to work with.

Concise

Headlines, captions, hooks

Use for short-form output: SMS, social options, quick variants.

Be direct. No preamble. Give 5–10 options. Each option should be short and usable.
Standard

Most business tasks

Normal marketing, writing, and decision tasks with enough structure to implement.

Use a practical, structured answer with enough detail to implement.
Detailed

Audits, strategy, SEO

Use for client-facing deliverables that need logic, trade-offs, and next steps.

Go deeper. Include assumptions, logic, tradeoffs, priority, examples, and execution steps.
Exhaustive

SOPs, full plans, investment docs

High-stakes deliverables that need edge cases, risks, and documentation.

Boil the ocean. Cover edge cases, risks, dependencies, tests, and documentation.
Exact Spec

Image & video production

Production-ready prompts for generative media tools.

Describe the visible scene, camera, lens, lighting, composition, motion, duration, aspect ratio, and negative prompts.
Use the Vault

Every workflow in the vault already has a model and brevity recommendation built in. Open the vault, pick a workflow, and the model choice is made for you.

Browse 1,096 workflows →