Claude Opus 4.5
VS
Claude Sonnet 4.6
Claude Opus 4.5 vs Claude Sonnet 4.6: which wins at real work?
22 task areas · same graded test runs · rank comparison only, so 0–100 and Elo collections never mix raw scores.
Claude Opus 4.5 and Claude Sonnet 4.6 split the 22 task areas we tested 11–11. Claude Sonnet 4.6 costs 1.7× less per token ($18 vs $30 per 1M).
Claude Sonnet 4.6 costs 1.7× less per token ($18 vs $30 per 1M).
Task by task
| Task area | Claude Opus 4.5 | Claude Sonnet 4.6 | Winner |
|---|---|---|---|
| Summarization & Meeting Notes |
#1
/ 110
Excellent
|
#37
/ 110
Excellent
|
Claude Opus 4.5 |
| Legal & HR |
#30
/ 110
Excellent
|
#3
/ 110
Excellent
|
Claude Sonnet 4.6 |
| Structured Output |
#82
/ 113
Strong
|
#105
/ 113
Usable
|
Claude Opus 4.5 |
| Frontend & Landing Pages |
#2
/ 109
Usable
|
#23
/ 109
Needs editing
|
Claude Opus 4.5 |
| RAG, Safety & Grounding |
#52
/ 113
Excellent
|
#73
/ 113
Excellent
|
Claude Opus 4.5 |
| Research & Competitive Analysis |
#17
/ 110
Excellent
|
#35
/ 110
Strong
|
Claude Opus 4.5 |
| Training & Education |
#24
/ 110
Excellent
|
#9
/ 110
Excellent
|
Claude Sonnet 4.6 |
| AI Strategy |
#16
/ 126
Strong
|
#3
/ 126
Strong
|
Claude Sonnet 4.6 |
| Knowledge & Docs |
#47
/ 110
Strong
|
#35
/ 110
Strong
|
Claude Sonnet 4.6 |
| Product & Project Management |
#60
/ 110
Strong
|
#48
/ 110
Strong
|
Claude Sonnet 4.6 |
| Creative & Comedy | #8 / 110 | #19 / 110 | Claude Opus 4.5 |
| Coding |
#39
/ 115
Strong
|
#29
/ 115
Strong
|
Claude Sonnet 4.6 |
| Content & Brand |
#21
/ 124
Strong
|
#31
/ 124
Strong
|
Claude Opus 4.5 |
| Presentations & Decks |
#45
/ 110
Excellent
|
#35
/ 110
Excellent
|
Claude Sonnet 4.6 |
| Sales |
#3
/ 110
Strong
|
#13
/ 110
Strong
|
Claude Opus 4.5 |
| Translation & Localization |
#5
/ 110
Excellent
|
#13
/ 110
Excellent
|
Claude Opus 4.5 |
| Chef / Home Cooking |
#14
/ 126
Strong
|
#20
/ 126
Usable
|
Claude Opus 4.5 |
| Customer Support |
#56
/ 113
Strong
|
#62
/ 113
Strong
|
Claude Opus 4.5 |
| Investor & Pitch |
#10
/ 66
Strong
|
#4
/ 66
Strong
|
Claude Sonnet 4.6 |
| Data & Analytics |
#9
/ 110
Excellent
|
#6
/ 110
Excellent
|
Claude Sonnet 4.6 |
| Executive Assistant |
#4
/ 112
Strong
|
#1
/ 112
Strong
|
Claude Sonnet 4.6 |
| Landing Pages |
#7
/ 72
Strong
|
#5
/ 72
Strong
|
Claude Sonnet 4.6 |
Rank = position among every model config we tested in that task area (lower is better). Sorted by biggest gap first.
Same task, both models — judged
Both models answered the same test case; an independent judge graded each. Quotes are the judge's actual rationale.
Frontend & Landing Pages
Lumen Health — compliance-careful healthtech (Landing Page Build)
“Copy is highly specific, benefit-led, and strictly adheres to compliance constraints without hype. Desktop design is clean and professional. However, the mobile layout is fundamentally broken with severe horizontal overflow across all sections.”
“vision judge failed: litellm.AuthenticationError: GeminiException - { "error": { "code": 400, "message": "API key not valid. Please pass a valid API key.", "status": "INVALID_ARGUMENT", "details": [ { "@type": "type.googleapis.com/google.rpc.ErrorInfo", "reason": "API_KEY_INVALID", "domain": "googleapis.com", "metadata": { "service": "generativelanguage.googleapis.com" } }, { "@type": "type.googleapis.com/google.rpc.LocalizedMessage", "locale": "en-US", "message": "API key not valid. Please pass a valid API key." } ] } } ”
Frequently asked
Is Claude Opus 4.5 better than Claude Sonnet 4.6?
They tie across the 22 task areas we benchmarked, at 11 apiece.
Which is cheaper, Claude Opus 4.5 or Claude Sonnet 4.6?
Claude Sonnet 4.6 costs 1.7× less per token ($18 vs $30 per 1M).
What is Claude Opus 4.5 better at?
Claude Opus 4.5 out-ranks Claude Sonnet 4.6 at Summarization & Meeting Notes, Structured Output, Frontend & Landing Pages.
What is Claude Sonnet 4.6 better at?
Claude Sonnet 4.6 out-ranks Claude Opus 4.5 at Legal & HR, Training & Education, AI Strategy.
More comparisons
This page is Spring Prompt, running in public
We just did this for every model. Do it for your prompt.
The rankings above come from running real tasks through real models and scoring every output. Spring Prompt is that same engine — pointed at your prompt, your test cases, and your definition of good.
- Generate test cases from your prompt — no eval set required to start.
- Compare models side by side with quality, cost and latency in one matrix.
- Optimise the winner until the scores say it's ready to ship.
Prompt × model results
12 test cases · 3 evals