I Tested Sonnet 5.5 vs Opus 5.5. What You Need to Know.

Product reviewed

Claude 5.5 models: Claude Sonnet 5.5 vs Claude Opus 5.5 (AI model comparison). The video focuses on speed, token usage, and real-world output quality/cost across multiple tasks.

---

Key features compared

• Speed / performance: how long each model took on the same (or equivalent) tasks.
• Cost: compared using token pricing and observed run costs.
• Output handling, including:
  • Web / landing page generation with interactive visuals
  • Motion design / showreels (including sound design)
  • Planning documents from vague prompts (research-backed action plans)
  • Investor decks + Excel sheets (including formula-heavy spreadsheets)
  • “Local AI models on my PC” explanations (skill-driven HTML explainer)
  • Interactive news dashboards pulling from X (live/refreshable UI)
  • Reference guide writing with defined quality rules (skill-driven)

---

Pricing & token costs (given explicitly)

• Sonnet 5.5
  • $2 per 1M input tokens
  • $10 per 1M output tokens
• Opus 5.5
  • $4 per 1M input tokens
  • $20 per 1M output tokens
• Pricing conclusion from the video: Opus is ~2× the cost of Sonnet (input and output).

---

Claude’s suggested usage (quoted in the video)

• Use Sonnet for:
  • “daily programming, bug fixing”
  • well-defined tasks with clear specification and verifiable results
• Use Opus 5.5 for:
  • complex work needing careful judgment
  • long-term agent programming
  • “maximum intelligence” for the hardest problems

Creator’s interpretation:

• Clear/objective tasks → Sonnet
• Vague/ambiguous tasks needing creativity / thinking partner → Opus

---

Task-by-task results

1) Landing page for “PerkForm” (same query, high effort)

Sonnet

• Better / more “correct” main section (the can/3D look considered more accurate)
• More effective “wow” in the opening section

Opus

• Opening section wasn’t as strong (“didn’t account for it” / less wow)
• Later storytelling / animations considered better overall
• Used a “fake image” approach (images requested later)

Performance / cost stats

• Time: Sonnet 28 min vs Opus 52 min
• Tokens: Sonnet 20M input / 134k output vs Opus 37M input / 148k output
• Cost: Opus ~2× as expensive

Verdict

• Sonnet “wins” (creator says Opus wasn’t ~2× better; recommends re-prompting for real images).

---

2) Motion design showreel for Glido (tight branding expectations)

Sonnet

• Creative; produced a showreel with generated motion/video
• Considered “pretty good” though not the best

Opus

• Dramatically better motion design with better taste in:
  • movement
  • sound design
• Pulled a quote from the website (research behavior)

Performance / cost stats

• Creator states Opus was faster, used fewer alpha tokens, but more expensive
• Cost: Sonnet $14.37 vs Opus $20.55

Verdict

• Opus wins (sound + movement quality emphasized).

---

3) “Vague” request: 90-day research-backed plan to reach 1,000 subscribers

Overall notes

• Both initially output a markdown file, then adjusted to be more readable.

Sonnet

• Produced an HTML plan that looked reasonable but was less aligned/clear to the creator

Opus

• More business-aligned Q4 sprint timeline (Oct 1 → Dec 29)
• Clearer breakdown by months/phases; easier to act on
• Framed by the creator as Opus being a better “thinking partner” for a fuzzy prompt

Performance / cost stats

• Time: Sonnet 14 min vs Opus 20 min
• Cost: Sonnet just under $5 vs Opus just over $9

Verdict

• Opus wins; creator says Opus leads 2-to-1 so far.

---

4) Investor deck + Excel spreadsheet from fictitious company data

Opus

• Investor presentation looked acceptable, but creator disliked the “infamous Claude font” vibe and wanted “more premium”
• Excel praised for formula coverage (many cell formulas; sheet updates dynamically)

Sonnet

• Presentation very similar to Opus (creator says “no difference” in feel)
• Excel looked “normal,” with no meaningful advantage; described as “might have given us too much of everything”

Verdict reasoning

• When output is practically identical, choose cheaper/faster

Verdict

• Opus wins because it was described as cheaper and faster with comparable output (despite general pricing expectations).

---

5) Skill-based task: “Local AI models that my PC can run” (HTML explainer)

Opus

• Clear rule explanations (e.g., 32GB rule, “must fit in your graphics card”)
• Better explanation of why gigabytes matter
• Slightly better reasoning clarity and alignment with creator expectations

Sonnet

• More visual and included extra explanations of “local” and storage concepts
• Creator felt it didn’t explain underlying reasoning as well

Cost/value angle

• With similar runtime, creator notes Sonnet costs about half
• Creator prefers Sonnet for value, assuming iteration via follow-up questions

Verdict

• Sonnet preferred (better cost/value despite Opus explaining reasoning better).

---

6) Interactive dashboard from X: “collect today’s news + interactive plan”

Sonnet dashboard

• Includes update tracking, trending/hot topics, marking relevance to the channel, pin/hide options, calendar views, and easy post opening
• Costs less, but creator felt Opus delivered a “better experience”

Opus dashboard

• More visual analytics (posts per hour, popularity gain/loss)
• Labels items by type (rumors/releases/discussions)
• Richer interaction for content planning

Performance / cost stated

• Opus ran ~7 minutes longer and cost $6 more

Verdict

• Creator says Opus is better, but under his cost rule Sonnet gets the win (Sonnet can likely reach similar results with better prompting/iteration).

---

7) Reference guide generation using a “quality definition” skill (YouTube-based)

Sonnet vs Opus

• Both produced structured guides with similar sections/length
• Creator says documents were extremely similar; chose the cheaper one

Performance / cost stats

• Sonnet about 2× as fast and $1.40 cheaper

Scores / overall result

• Creator concludes: Sonnet wins
• Final tally mentioned: 4 in favor of Sonnet vs 3 in favor of Opus (across these sessions)

Verdict

• Sonnet wins due to equivalently good quality at lower cost/speed.

---

Overall outcomes (video’s final take)

• Opus generally wins when tasks are vague, require deep judgment, and benefit from creative thinking (especially motion/sound design and structured planning from ambiguity).
• Sonnet generally wins when tasks are clear, spec-driven, verifiable, or when cost matters and you can iterate with follow-up prompts.

Creator’s overarching decision rule

• Don’t argue universally “which model is better.”
• Choose based on task clarity and need for judgment/creativity.

Final recommendation (implied by creator)

• Start with Sonnet for well-defined goals and fast/cost-effective iterations.
• Use Opus when you want stronger taste, higher-quality creative judgment, and better handling of ambiguity.

---

Unique pros mentioned (per model)

Sonnet 5.5 — Pros

• Faster in several tests (e.g., landing page, reference guide)
• Much cheaper (explicit token pricing; multiple task costs favor Sonnet)
• Strong for tasks with clear specification and verifiable outcomes
• In at least one case, better opening/main section (landing page)
• Preferred on value-focused tasks (local model explanation cost/value; dashboard “win” by cost rule; reference guide cheaper and similar quality)

Opus 5.5 — Pros

• Better at complex creative work and “thinking partner” behavior for vague requests
• Stronger motion design and sound design (repeatedly highlighted)
• Better dashboard experience and richer interactive/visual analytics
• Produces formula-heavy spreadsheets (praised for maintainability)
• Often yields more “complete/aligned” outputs (e.g., Q4 timeline plan)

---

Unique cons mentioned (per model)

Sonnet 5.5 — Cons

• Doesn’t “think as deeply” for ambiguous tasks (may comply, but not go as far as Opus)
• Some outputs can feel less premium on quality/behavior (e.g., some showreel aspects; less strong dashboard visuals/analytics vs Opus)

Opus 5.5 — Cons

• Typically more expensive and sometimes slower
• Can produce less accurate or less preferred elements unless guided (e.g., “fake image” issue)
• Creator felt costs aren’t justified when improvements aren’t proportionally better

---

Comparisons explicitly made

• Opus vs Sonnet cost ratio: Opus about 2× pricing; runtime/cost often followed that, but quality scaling wasn’t always 2×.
• When quality differences are small: creator prefers cheaper/faster (investor deck; reference guide).
• When prompts are fuzzy: Opus is more likely to outperform Sonnet (90-day plan).

---

Verdict / recommendation

Choose Claude Sonnet 5.5 by default for:

• clear, spec-driven, verifiable tasks
• lower cost and faster iteration

Choose Claude Opus 5.5 when:

• you need best-in-class creative judgment
• especially for motion/audio quality
• the prompt is vague/ambiguous and needs a “thinking partner” to structure decisions

---

Speakers / viewpoints

Only one main speaker is present throughout; no distinct multi-speaker perspectives were clearly separated in the subtitles.