Claude Sonnet 4.5
VS
Kimi K2.7 Code
Claude Sonnet 4.5 vs Kimi K2.7 Code: 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.
Kimi K2.7 Code wins 18 of 22 task areas we tested; Claude Sonnet 4.5 takes 4. Kimi K2.7 Code costs 4.2× less per token ($4.24 vs $18 per 1M).
Kimi K2.7 Code costs 4.2× less per token ($4.24 vs $18 per 1M).
Task by task
| Task area | Claude Sonnet 4.5 | Kimi K2.7 Code | Winner |
|---|---|---|---|
| Product & Project Management |
#71
/ 110
Strong
|
#10
/ 110
Excellent
|
Kimi K2.7 Code |
| Legal & HR |
#69
/ 110
Strong
|
#14
/ 110
Excellent
|
Kimi K2.7 Code |
| Summarization & Meeting Notes |
#2
/ 110
Excellent
|
#47
/ 110
Excellent
|
Claude Sonnet 4.5 |
| Presentations & Decks |
#43
/ 110
Excellent
|
#5
/ 110
Excellent
|
Kimi K2.7 Code |
| Sales |
#10
/ 110
Strong
|
#48
/ 110
Usable
|
Claude Sonnet 4.5 |
| Structured Output |
#80
/ 113
Strong
|
#46
/ 113
Excellent
|
Kimi K2.7 Code |
| Customer Support |
#43
/ 113
Strong
|
#10
/ 113
Strong
|
Kimi K2.7 Code |
| Investor & Pitch |
#42
/ 66
Usable
|
#15
/ 66
Strong
|
Kimi K2.7 Code |
| Content & Brand |
#74
/ 124
Usable
|
#48
/ 124
Strong
|
Kimi K2.7 Code |
| Executive Assistant |
#31
/ 112
Strong
|
#6
/ 112
Strong
|
Kimi K2.7 Code |
| Frontend & Landing Pages |
#29
/ 109
Needs editing
|
#54
/ 109
Needs editing
|
Claude Sonnet 4.5 |
| AI Strategy |
#67
/ 126
Strong
|
#44
/ 126
Strong
|
Kimi K2.7 Code |
| Translation & Localization |
#25
/ 110
Excellent
|
#48
/ 110
Excellent
|
Claude Sonnet 4.5 |
| Chef / Home Cooking |
#80
/ 126
Usable
|
#58
/ 126
Usable
|
Kimi K2.7 Code |
| Creative & Comedy | #66 / 110 | #44 / 110 | Kimi K2.7 Code |
| Coding |
#62
/ 115
Strong
|
#41
/ 115
Strong
|
Kimi K2.7 Code |
| Research & Competitive Analysis |
#39
/ 110
Strong
|
#20
/ 110
Excellent
|
Kimi K2.7 Code |
| RAG, Safety & Grounding |
#39
/ 113
Excellent
|
#21
/ 113
Excellent
|
Kimi K2.7 Code |
| Landing Pages |
#30
/ 72
Strong
|
#17
/ 72
Strong
|
Kimi K2.7 Code |
| Knowledge & Docs |
#28
/ 110
Strong
|
#16
/ 110
Excellent
|
Kimi K2.7 Code |
| Training & Education |
#51
/ 110
Excellent
|
#39
/ 110
Excellent
|
Kimi K2.7 Code |
| Data & Analytics |
#33
/ 110
Excellent
|
#31
/ 110
Excellent
|
Kimi K2.7 Code |
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.
Presentations & Decks
Action titles from findings (Tradewinds) (Action Titles)
“The response fails to use an Answer-First deductive structure and lacks a clear ask or next steps.”
“The response perfectly executes the prompt's requirements. It leads with strong, active-voice recommendations and builds a coherent, answer-first narrative across the three slides. The titles are excellent examples of executive-level action titles, and the supporting notes provide precise, non-redundant context.”
Frequently asked
Is Claude Sonnet 4.5 better than Kimi K2.7 Code?
Across 22 task areas we benchmarked, Kimi K2.7 Code ranks higher in 18 and Claude Sonnet 4.5 in 4.
Which is cheaper, Claude Sonnet 4.5 or Kimi K2.7 Code?
Kimi K2.7 Code costs 4.2× less per token ($4.24 vs $18 per 1M).
What is Claude Sonnet 4.5 better at?
Claude Sonnet 4.5 out-ranks Kimi K2.7 Code at Summarization & Meeting Notes, Sales, Frontend & Landing Pages.
What is Kimi K2.7 Code better at?
Kimi K2.7 Code out-ranks Claude Sonnet 4.5 at Product & Project Management, Legal & HR, Presentations & Decks.
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