GENERAL LLMs (Modules 10–14)

For LLMs the VS field compares to the nearest peer, not Claude Code (which is a coding harness, not a chat LLM). Your reflex baseline here is "Claude in the chat app" — these are measured against that and each other. Prices verified June 2026.

MODULE 10: ChatGPT / GPT-5.5 CATEGORY: LLM DEPTH: CORE ONE-LINE SUMMARY: The default general-purpose AI — strongest ecosystem (voice, agent mode, Canvas, custom GPTs, Codex, image gen via DALL-E), broadest mindshare. VS NEAREST PEER (Claude / Gemini): Widest feature surface and the name clients actually say. vs Claude: more product features (voice, agent mode, GPT store) but you may find Claude stronger for long-form coding judgment. vs Gemini: smaller context window but a more mature agent/tool ecosystem. BEST FOR:

• The "talk to clients in their language" tool — everyone knows ChatGPT.
• Voice brainstorming, agent mode tasks, and DALL-E image gen without leaving chat.
• Custom GPTs as cheap productized assistants you can build and (soon) sell. WEAKNESS: Smaller context window than Gemini; the tier sprawl (Free/Go/Plus/Pro-$100/Pro-$200) is confusing; Pro tier is expensive for marginal gains for most solo builders. COST: ⚠ verify — Free; Go $8/mo; Plus $20/mo (GPT-5.5 routing, agent mode, 10 deep research/mo); Pro $100/mo (5× limits) and Pro $200/mo (GPT-5.5 Pro, 20× limits, ~1M context, 250 deep research). Business $25/user/mo. HANDS-ON TASK: In Plus, build a Custom GPT seeded with your freelance service description + FAQ, so it answers prospect questions in your voice. That's a sellable artifact in 15 min. GOTCHA: Five overlapping paid tiers means clients/students conflate them. Know exactly which features gate behind Plus vs the two Pro tiers before you advise anyone to upgrade.

MODULE 11: Gemini 3.x (Pro / Flash) CATEGORY: LLM DEPTH: CORE ONE-LINE SUMMARY: Google's frontier model line — enormous 1M-token context, native multimodal (image/video/audio), and deep Google Workspace + Android integration. VS NEAREST PEER (ChatGPT): Far bigger context window (1M) → drop whole codebases/PDFs/videos in at once. Flash models are extremely cheap+fast for API/automation. vs ChatGPT: weaker third-party plugin ecosystem but better raw long-context + cheaper bulk inference. BEST FOR:

• Long-context jobs: dump an entire repo, a 300-page PDF, or a long video and ask across all of it.
• Cheap, fast API automation (Gemini 3.5 Flash) inside n8n/FastAPI pipelines.
• Anyone living in Gmail/Docs/Sheets — it's wired into Workspace. WEAKNESS: Product surface changes fast and confusingly (AI Plus/Pro/Ultra + Google One bundling); third-party ecosystem thinner than OpenAI's; output can be more cautious. COST: ⚠ verify — Free tier; Google AI Plus $7.99/mo; AI Pro $19.99/mo (Gemini 3.x Pro, 1M context, 2TB storage); AI Ultra dropped to ~$99.99–200/mo after I/O 2026. API: Gemini 3.5 Flash ~$1.50/$9.00 per M tokens (in/out). HANDS-ON TASK: Paste an entire small repo (or a long client PDF) into Gemini's 1M-context window and ask for an architecture summary + risks. Feel what 1M context unlocks vs. chunking. GOTCHA: The consumer plan names (AI Plus/Pro/Ultra) and the API model names (3.5 Flash/Pro) are different products with different pricing. Don't quote one when you mean the other.

MODULE 12: Grok (xAI) CATEGORY: LLM DEPTH: REFERENCE ONE-LINE SUMMARY: xAI's model with live access to the X/Twitter firehose, a less-filtered tone, and strong real-time/news awareness. VS NEAREST PEER (ChatGPT/Perplexity): Its edge is real-time X data and an unfiltered personality. For coding/general reasoning it's competitive but not the leader. Think "ChatGPT with a live social-media nervous system and fewer guardrails." BEST FOR:

• Real-time pulse on breaking news, trends, and X discourse (great for content/marketing research).
• Tasks where you want a blunter, less hedged tone.
• Anyone already paying for X Premium+ (Grok comes bundled). WEAKNESS: Smaller ecosystem; quality/consistency more variable; "less filtered" can mean less reliable. Tied closely to the X platform. COST: ⚠ verify — Free (limited); via X: X Premium $8/mo, Premium+ $40/mo; standalone SuperGrok Lite $10/mo, SuperGrok $30/mo, SuperGrok Heavy $300/mo. API: Grok 4.x ~$1.25–2/$2.50–6 per M tokens. COMMUNICATION SHORTCUT: "Grok's differentiator is live X data and a looser tone — I reach for it for real-time trend/news research, not as my main reasoning model." GOTCHA: "Real-time from X" means it can confidently surface unverified social-media claims. Treat its live-data answers as leads, not facts.

MODULE 13: DeepSeek CATEGORY: LLM DEPTH: SKIM ONE-LINE SUMMARY: An open-weight Chinese model line with frontier-ish reasoning at a fraction of the API cost — and a free consumer chat. VS NEAREST PEER (GPT/Gemini API): The cost story. DeepSeek API is ~10–100× cheaper than premium US models, and weights are open so you can self-host. Quality is strong for the price, especially on reasoning/coding. Trade-off: data-residency/compliance concerns (Chinese hosting) and slightly behind frontier on the hardest tasks. BEST FOR:

• Bulk/automation LLM calls where cost dominates (classification, extraction, batch edits via Aider).
• Self-hosting an open-weight model for data-sensitive client work.
• Cheap reasoning in your FastAPI/n8n pipelines. WEAKNESS: Data-privacy/compliance optics for some clients (hosted in China); occasional content restrictions; not the absolute frontier on the hardest problems. COST: ⚠ verify — Consumer chat FREE; 5M free API tokens on signup. API ~$0.14/M input, $0.28/M output (V4 Flash); cache hits ~$0.0028/M (huge savings on repeated context). HANDS-ON TASK: Sign up, grab the free tokens, and point Aider (Module 9) at DeepSeek to do a repo refactor — measure the cost vs. doing it on a premium model. The delta is the pitch. GOTCHA: For client work, the "where is the data processed" question is real. Use the open weights self-hosted (or a Western host) when a client has data-residency requirements; don't pipe their data to the default API blindly.

MODULE 14: Perplexity AI CATEGORY: LLM DEPTH: SKIM ONE-LINE SUMMARY: An answer engine — every response is web-grounded with inline citations. It's "Google + LLM" rather than a raw chatbot. VS NEAREST PEER (ChatGPT/Gemini): Optimized for cited, current, factual answers, not creativity or coding. Where ChatGPT might hallucinate a fact, Perplexity shows its sources. Weaker as a general assistant; stronger as a research tool. BEST FOR:

• Fast, sourced research before writing/quoting (e.g., verifying tool pricing for this course).
• Competitive/market research for client proposals with citations attached.
• Replacing a dozen Google tabs when you need an answer + receipts. WEAKNESS: Not for deep coding or long creative work; answer quality depends on what's indexed; the value-add shrinks as base models add their own web search. COST: ⚠ verify — Free; Pro $20/mo ($200/yr, unlimited searches, full model choice, $5/mo API credits); Max $200/mo; Education Pro $10/mo (student); Enterprise from $40/seat. HANDS-ON TASK: Ask Perplexity to compare two competing tools in your stack with sources, then click through the citations to confirm. Notice how the citation trail makes it safe to quote. GOTCHA: Citations create false confidence — Perplexity sometimes cites a source that doesn't actually support the claim. Click through before you repeat it in paid work.

LLMs — at a glance

Sources (verified June 2026)

• ChatGPT: cloudzero
• Gemini: screenapp, Google blog I/O 2026
• Grok: felloai
• DeepSeek: felloai, official
• Perplexity: screenapp

Next: 03-image.md.