Claude (Anthropic)

Best for

• Drafting client communications that require warmth + regulatory awareness
• Long-document summarization (10-K filings, prospectuses, court rulings — Opus handles up to 1M tokens)
• Multi-step analytical reasoning where you want the model to show its work
• Code-review and Excel formula explanation where precision matters more than speed

Avoid

• Real-time data — Claude is knowledge-cutoff bound. Use Perplexity for current news.
• Image generation — Claude doesn't generate images. Use DALL·E or Gemini.
• Bulk operations where you'd benefit from API automation — Claude's API is fine but ChatGPT's ecosystem is broader.

Shortcuts

• Use Projects (Pro+) to load firm-specific context once — disclaimers, compliance rules, voice samples — and reference it across many chats.
• Use Artifacts for any document you'll iterate on (emails, summaries, code) — they render cleanly and accept revision requests in-line.
• Custom Styles let you save your firm's voice so every output sounds consistent across the team.
• For research analysis, paste the document first and ask the question second — Claude attends better to recent context.

Model variants

Opus 4.7 (1M context) — best for long-document analysis and complex reasoning. Sonnet 4.6 — the daily driver, 80% of the quality at 20% of the cost. Haiku 4.5 — speed/cost-optimized, fine for quick formatting and short tasks.

Compliance gotchas

Pricing

Free tier limited (Sonnet only, message limits). Pro $20/mo (full Sonnet + Opus access). Team $30/user/mo (recommended for advisory firms — adds admin controls, zero retention, and Project sharing). Enterprise (contact sales) for SSO + audit.