Compare / DeepSeek-V4.1-Flash-1017 vs Mercury 3

DeepSeek-V4.1-Flash-1017 vs Mercury 3

Which is cheaper, which is more capable, and which does better at each kind of work, from 53 results on 4 sources that measured both models.

DeepSeekDeepSeek-V4.1-Flash-1017 Released 17 Oct 2026 · $0.30 in · $1.20 out
InceptionMercury 3 Released 19 Oct 2026 · $0.20 in · $0.70 out
DeepSeek-V4.1-Flash-1017 wins 338 too close to callMercury 3 wins 12

Of the 53 results both models have, the better value on each. Results whose reported ranges overlap are too close to call.

Results as of 20 October 2026 (catalogue release 2026-10-20-96eb40c3ec63).

DeepSeek-V4.1-Flash-1017 or Mercury 3? The short answer

They score about the same, so pick Mercury 3: it costs 2.2 times less per deck on DeckBench.

  • Spring Prompt overall: DeepSeek-V4.1-Flash-1017 44, Mercury 3 39 (out of 100; how the score works).
  • Mercury 3 is cheaper to run: $0.012 against $0.026 per deck on DeckBench (2.2 times less).
  • Mercury 3 is newer: released 19 Oct 2026, against 17 Oct 2026 for DeepSeek-V4.1-Flash-1017.
  • On our own benchmarks, on the results both models have: DeepSeek-V4.1-Flash-1017 does better for decks from an analysis, user surveys and marketing planning.

Head to head

44
Spring Prompt overall
39
$0.52
Price per million tokens, blendedlower is better
$0.33
1,048,576
Context window, tokens
1,000,000

Every fact side by side

DeepSeek-V4.1-Flash-1017Mercury 3
Price per million tokens$0.30 in · $1.20 out$0.20 in · $0.70 out
Blended (3 in : 1 out)$0.52$0.33
Cheapest host, blended––
Cost per deck (DeckBench)$0.026$0.012
Cached input, per million$0.006 cached input$0.020 cached input
Intelligence Index45.1–
Spring Prompt overall4439
Output speed, tokens a second212–
Context window1,048,576 tokens1,000,000 tokens
Released17 Oct 202619 Oct 2026 (newer)
WeightsOpenClosed
Developer based inChinathe US

Prices are the developer's list price where we have read it, otherwise the typical host on OpenRouter; see every model's price. Intelligence and speed from Artificial Analysis.

Which is better for what

Each model's rank among the models on sale today, for each kind of work, from our own benchmarks and licensed sources. Further right is better; the better rank is in green.

Product listingsTurning a sparse product feed and photos into listings that can go live 21 of 24 not measured
Decks from an analysisTurning a finished analysis into a deck you could present as it is 12 of 24 18 of 24
User surveysPlanning a user survey and reading its results without being misled 5 of 13 9 of 13
Marketing planningPlanning a year of ad spend without overspending 11 of 24 18 of 24
Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls 22 of 168 not measured
Reasoning and knowledgeHard questions across science, maths and general knowledge 22 of 202 not measured
Speed under pressureGood decisions against a real clock (fast chess) 21 of 27 10 of 27

From our research

Every shared result

Each model's best result on every measure both have, grouped by benchmark, with the setting that produced it. Under each measure, every model that benchmark has measured is a grey tick, best to the right, so you can see where the two sit in the field. The headline measures of each benchmark come first; the rest are under "Show all". Green marks the better value; where a source reports ranges that overlap, neither is marked.

MeasureDeepSeek-V4.1-Flash-1017Mercury 3
BulletBench measured by Spring Prompt · 3 results
Cost per game · Bullet 60sUS dollars, lower is better 13 · 7 of 32 measured $0.0047low reasoning $0.0020no reasoning
Ladder Elo · Bullet 60sladder Elo, higher is better 27 · 15 of 32 measured 147low reasoning 491no reasoning
Games lost on time · Bullet 60s% of games, lower is better 29 · 1 of 32 measured 87.5%low reasoning 0.0%no reasoning
DeckBench measured by Spring Prompt · 3 results
Cost per deckUS dollars, lower is better 7 · 4 of 24 measured $0.0255 $0.0116
Deck ratingrating, higher is betterThe rating weighs each win by how strong the other deck was; the win rate counts wins against whichever decks a model happened to meet, so the two can disagree. 12 · 18 of 24 measured 1,145 812
Accurate decks% of tasks, higher is better 14 · 20 of 24 measured 33.3% 0.0%
ROASBench measured by Spring Prompt · 3 results
Overall scorescore out of 100, higher is better 11 · 18 of 24 measured 41.4 26.1
Audience scorescore out of 100, higher is better 9 · 21 of 24 measured 65.3 43.2
Business scorescore out of 100, higher is better 11 · 18 of 24 measured 38.3 17.6
SurveyBench measured by Spring Prompt · 3 results
Analysis ratingrating, higher is betterThe rating weighs each win by how strong the other analysis was; the win rate counts wins against whichever analyses a model happened to meet, so the two can disagree. 5 · 9 of 13 measured 1,221 752
Cost per taskUS dollars, lower is better 4 · 3 of 12 measured $0.0373 $0.0216
Analysis scorepoints of 100, higher is better 4 · 12 of 13 measured 90.2 42.2
Show all 53 results (41 more)
MeasureDeepSeek-V4.1-Flash-1017Mercury 3
BulletBench measured by Spring Prompt · 2 results
Invalid moves · Bullet 60s% of moves, lower is better 1 · 1 of 32 measured 0.0%low reasoning 0.0%low reasoning
Median move time · Bullet 60smilliseconds, lower is better 31 · 3 of 32 measured 3.3 slow reasoning 0.4 sno reasoning
DeckBench measured by Spring Prompt · 15 results
Caveat dropped% of tasks, lower is better 12 · 12 of 24 measured 16.7% 16.7%
Clean layout% of tasks, higher is better 20 · 6 of 24 measured 16.7% 66.7%
Design quality% of the maximum, higher is better 9 · 16 of 24 measured 68.5% 58.5%
Draft figure quoted% of tasks, lower is better 19 · 12 of 24 measured 50.0% 33.3%
Findings missing% of tasks, lower is better 1 · 17 of 24 measured 0.0% 16.7%
Head-to-head win rate% of comparisons, higher is betterThe rating weighs each win by how strong the other deck was; the win rate counts wins against whichever decks a model happened to meet, so the two can disagree. 14 · 16 of 24 measured 53.3% 44.0%
Layout defects% of tasks, lower is better 20 · 6 of 24 measured 83.3% 33.3%
Misleading metric used% of tasks, lower is better 1 · 22 of 24 measured 0.0% 16.7%
Output tokens per decktokens, lower is better 1 · 2 of 3 measured 21,019 32,448
Presentable decks% of tasks, higher is better 12 · 12 of 24 measured 0.0% 0.0%
Recommendation late or wrong% of tasks, lower is better 9 · 22 of 24 measured 16.7% 83.3%
Slides needing work% of tasks, lower is better 16 · 7 of 24 measured 66.7% 33.3%
Time per deckseconds, lower is better 7 · 9 of 24 measured 73.0 83.3
Unsupported claims% of tasks, lower is better 11 · 11 of 24 measured 16.7% 16.7%
Unsupported numbers% of tasks, lower is better 1 · 20 of 24 measured 0.0% 16.7%
ROASBench measured by Spring Prompt · 6 results
Consistency scorescore out of 100, higher is better 15 · 19 of 24 measured 20.2 15.1
Contribution profitsimulated US dollars, higher is better 10 · 18 of 24 measured $585,572 $104,868
Cost of a runUS dollars, lower is better 8 · 6 of 24 measured $0.0799 $0.0533
Months over budgetmonths of 12, lower is better 1 · 1 of 24 measured 0 0
Planning scorescore out of 100, higher is better 20 · 23 of 24 measured 54.3 53.8
Return on ad spendprofit per $1 spent, higher is better 12 · 18 of 24 measured $0.44 $0.10
SurveyBench measured by Spring Prompt · 18 results
Head-to-head win rate% of comparisons, higher is betterThe rating weighs each win by how strong the other analysis was; the win rate counts wins against whichever analyses a model happened to meet, so the two can disagree. 5 · 6 of 13 measured 60.2% 48.6%
Leading or double-barrelled questions% of tasks, lower is better 3 · 8 of 13 measured 11.1% 55.6%
Mistaken next steps% of tasks, lower is better 1 · 5 of 13 measured 0.0% 22.2%
Numbers right% of answers, higher is better 1 · 13 of 13 measured 100.0% 63.3%
Objectives unanswerable% of tasks, lower is better 7 · 1 of 13 measured 11.1% 0.0%
Output tokens per tasktokens, lower is better 30,893 57,343
Precision ignored% of tasks, lower is better 1 · 1 of 13 measured 0.0% 0.0%
Research scorepoints of 100, higher is better 4 · 11 of 13 measured 90.9 56.8
Sound research% of tasks, higher is better 3 · 6 of 13 measured 33.3% 0.0%
Stakeholder's view presumed% of tasks, lower is better 1 · 7 of 13 measured 0.0% 22.2%
Survey plan scorepoints of 100, higher is better 4 · 6 of 13 measured 92.5 90.8
Survey structure% of tasks, lower is better 10 · 7 of 13 measured 33.3% 22.2%
Time per taskseconds, lower is better 6 · 10 of 12 measured 134 194
Traps handled% of traps, higher is better 5 · 11 of 13 measured 91.3% 60.7%
Traps missed% of tasks, lower is better 5 · 7 of 13 measured 44.4% 88.9%
Unsupported findings% of tasks, lower is better 4 · 4 of 13 measured 44.4% 44.4%
Wrong numbers% of tasks, lower is better 1 · 9 of 13 measured 0.0% 44.4%
Wrong recommendation% of tasks, lower is better 3 · 12 of 13 measured 11.1% 77.8%

Quick answers

Is DeepSeek-V4.1-Flash-1017 better than Mercury 3?

They score about the same, so pick Mercury 3: it costs 2.2 times less per deck on DeckBench.

Which is cheaper, DeepSeek-V4.1-Flash-1017 or Mercury 3?

Mercury 3 is cheaper to run: $0.012 against $0.026 per deck on DeckBench (2.2 times less).

Which is better for decks from an analysis, DeepSeek-V4.1-Flash-1017 or Mercury 3?

DeepSeek-V4.1-Flash-1017 ranks 12 of 24 models on sale for turning a finished analysis into a deck you could present as it is, against 18 for Mercury 3.

Which is better for user surveys, DeepSeek-V4.1-Flash-1017 or Mercury 3?

DeepSeek-V4.1-Flash-1017 ranks 5 of 13 models on sale for planning a user survey and reading its results without being misled, against 9 for Mercury 3.

Which is better for marketing planning, DeepSeek-V4.1-Flash-1017 or Mercury 3?

DeepSeek-V4.1-Flash-1017 ranks 11 of 24 models on sale for planning a year of ad spend without overspending, against 18 for Mercury 3.

Which is better for speed under pressure, DeepSeek-V4.1-Flash-1017 or Mercury 3?

Mercury 3 ranks 10 of 27 models on sale for good decisions against a real clock (fast chess), against 21 for DeepSeek-V4.1-Flash-1017.

Which has the bigger context window, DeepSeek-V4.1-Flash-1017 or Mercury 3?

DeepSeek-V4.1-Flash-1017: 1,048,576 tokens, against 1,000,000 for Mercury 3.

Run DeepSeek-V4.1-Flash-1017 and Mercury 3 on your own prompt

Benchmarks aren't your data. Try both side by side in the playground, with the cost of every answer.

Try the playgroundTalk to us