DeepSeek V4 Pro vs Gemini 3.1 Flash Lite: 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.
DeepSeek V4 Pro wins 18 of 22 task areas we tested; Gemini 3.1 Flash Lite takes 4. DeepSeek V4 Pro costs 1.3× less per token ($1.305 vs $1.75 per 1M).
DeepSeek V4 Pro costs 1.3× less per token ($1.305 vs $1.75 per 1M).
Task by task
| Task area | DeepSeek V4 Pro | Gemini 3.1 Flash Lite | Winner |
|---|---|---|---|
| Executive Assistant |
#38
/ 112
Strong
|
#101
/ 112
Usable
|
DeepSeek V4 Pro |
| Presentations & Decks |
#31
/ 110
Excellent
|
#80
/ 110
Strong
|
DeepSeek V4 Pro |
| Legal & HR |
#40
/ 110
Excellent
|
#86
/ 110
Strong
|
DeepSeek V4 Pro |
| Summarization & Meeting Notes |
#55
/ 110
Excellent
|
#13
/ 110
Excellent
|
Gemini 3.1 Flash Lite |
| Chef / Home Cooking |
#32
/ 126
Usable
|
#70
/ 126
Usable
|
DeepSeek V4 Pro |
| Creative & Comedy | #50 / 110 | #85 / 110 | DeepSeek V4 Pro |
| RAG, Safety & Grounding |
#27
/ 113
Excellent
|
#58
/ 113
Excellent
|
DeepSeek V4 Pro |
| Landing Pages |
#16
/ 72
Strong
|
#43
/ 72
Strong
|
DeepSeek V4 Pro |
| Sales |
#44
/ 110
Usable
|
#67
/ 110
Usable
|
DeepSeek V4 Pro |
| Product & Project Management |
#12
/ 110
Excellent
|
#29
/ 110
Excellent
|
DeepSeek V4 Pro |
| Data & Analytics |
#60
/ 110
Excellent
|
#44
/ 110
Excellent
|
Gemini 3.1 Flash Lite |
| Coding |
#43
/ 115
Strong
|
#58
/ 115
Strong
|
DeepSeek V4 Pro |
| AI Strategy |
#68
/ 126
Strong
|
#54
/ 126
Strong
|
Gemini 3.1 Flash Lite |
| Content & Brand |
#29
/ 124
Strong
|
#43
/ 124
Strong
|
DeepSeek V4 Pro |
| Structured Output |
#23
/ 113
Excellent
|
#36
/ 113
Excellent
|
DeepSeek V4 Pro |
| Research & Competitive Analysis |
#44
/ 110
Strong
|
#56
/ 110
Usable
|
DeepSeek V4 Pro |
| Investor & Pitch |
#33
/ 66
Strong
|
#44
/ 66
Usable
|
DeepSeek V4 Pro |
| Knowledge & Docs |
#25
/ 110
Strong
|
#33
/ 110
Strong
|
DeepSeek V4 Pro |
| Training & Education |
#41
/ 110
Excellent
|
#33
/ 110
Excellent
|
Gemini 3.1 Flash Lite |
| Translation & Localization |
#32
/ 110
Excellent
|
#39
/ 110
Excellent
|
DeepSeek V4 Pro |
| Frontend & Landing Pages |
#6
/ 109
Usable
|
#10
/ 109
Usable
|
DeepSeek V4 Pro |
| Customer Support |
#52
/ 113
Strong
|
#54
/ 113
Strong
|
DeepSeek V4 Pro |
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.
Executive Assistant
Angry contractor follow-up (Tactful Rewrite Test)
“The response perfectly executes the prompt's instructions. It maintains a firm, clear, and professional tone while entirely removing the emotional heat of the original message. It meets all constraints, preserves all required information, and is ready to send with only a minor placeholder adjustment.”
“The model failed the hard constraint to keep the output under 120 words by providing multiple options. Furthermore, one option contains a glaring typo, and another remains slightly aggressive, undermining the 'calm' requirement.”
Legal & HR
Data-processing condition (Lumen) (Plain-English Explainer)
“The model accurately translates the clause into plain English and preserves all material conditions, including the nested exceptions. However, it slightly over-extrapolates by stating the processor must 'wait for your response' (the clause only requires informing before processing) and lacks a standard legal disclaimer.”
“The model perfectly preserves all material conditions, exceptions, and nested exceptions in clear, plain English without flattening the legal meaning. It includes an excellent disclaimer. The only minor flaw is a slight misphrasing in one transition sentence ('conditions you must follow' instead of 'conditions the vendor must follow'), though the bullet points correctly assign the obligations to the vendor. Mentioning HIPAA/GDPR as examples of privacy laws is helpful context for a clinic administrator and does not constitute fabricated authority.”
Frequently asked
Is DeepSeek V4 Pro better than Gemini 3.1 Flash Lite?
Across 22 task areas we benchmarked, DeepSeek V4 Pro ranks higher in 18 and Gemini 3.1 Flash Lite in 4.
Which is cheaper, DeepSeek V4 Pro or Gemini 3.1 Flash Lite?
DeepSeek V4 Pro costs 1.3× less per token ($1.305 vs $1.75 per 1M).
What is DeepSeek V4 Pro better at?
DeepSeek V4 Pro out-ranks Gemini 3.1 Flash Lite at Executive Assistant, Presentations & Decks, Legal & HR.
What is Gemini 3.1 Flash Lite better at?
Gemini 3.1 Flash Lite out-ranks DeepSeek V4 Pro at Summarization & Meeting Notes, Data & Analytics, 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