DeepSeek-V3 vs Ministral 3 14B
Wins 5 of 6 areas
Coding · Reasoning · Facts · Long documents · Following instructions
Wins 1 of 6 areas
Agents
DeepSeek-V3 is the stronger all-rounder.Ministral 3 14B is cheaper and better at agents.
Scores updated · 26 tests both models report · How we compare
Where each one wins
Tests won in each of the six areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- CodingWriting and fixing software30DeepSeek-V33 of 3 tests
- FactsGetting facts right instead of making them up30DeepSeek-V33 of 3 tests
- ReasoningHard problems that need careful thinking10DeepSeek-V31 of 3 tests · 2 ties
- Long documentsFinding answers in very long texts10DeepSeek-V31 of 1 test
- Following instructionsDoing exactly what it is asked10DeepSeek-V31 of 1 test
- AgentsCarrying out multi-step tasks on its own01Ministral 3 14B1 of 3 tests · 2 ties
Long documents and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
Ministral 3 14B costs 65% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where DeepSeek-V3 pulls ahead
- Code for real scientific research problemsSciCode+15.2points ahead
- Reasons across sets of long documentsAA-LCR+14.4points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+11.9points ahead
Where Ministral 3 14B pulls ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+1.9points ahead
Every test, side by side
All 26 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V3
- SciCodeDeepSeek-V3 by 15.23923.8+15.2
- Terminal-Bench HardDeepSeek-V3 by 10.715.24.5+10.7
- Terminal-Bench 2.1DeepSeek-V3 by 7.216.99.7+7.2
AgentsMinistral 3 14B
- τ-Bench V3 · BankingMinistral 3 14B by 1.94.76.6+1.9
- GDPValtie00tie
- Terminal-Bench 4.0tie00tie
ReasoningDeepSeek-V3
- GPQA DiamondDeepSeek-V3 by 8.365.557.2+8.3
- Humanity's Last Examtie4.74.6tie
- CritPttie00tie
FactsDeepSeek-V3
- Vectara HHEM hallucination ratelower is betterDeepSeek-V3 by 13.36.119.4+13.3
- AA-Omniscience · AccuracyDeepSeek-V3 by 11.925.413.6+11.9
- AA-Omniscience · Non-hallucinationDeepSeek-V3 by 6.614.17.5+6.6
Other results12 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- AIME 2024Ministral 3 14B by 50.639.289.8+50.6
- Arena HardDeepSeek-V3 by 36.391.455.1+36.3
- AIME 2025Ministral 3 14B by 33.751.385+33.7
- AA-OmniscienceDeepSeek-V3 by 25.7-40.7-66.4+25.7
- τ²-Bench Telecom (AA run)DeepSeek-V3 by 19.947.127.2+19.9
- LiveCodeBenchMinistral 3 14B by 15.449.264.6+15.4
- vectara_factual_consistencyDeepSeek-V3 by 13.393.980.6+13.3
- MMLUDeepSeek-V3 by 9.188.579.4+9.1
- Artificial Analysis Coding IndexDeepSeek-V3 by 8.62314.4+8.6
- vectara_avg_summary_lengthMinistral 3 14B by 54.1 rating points81.7135.8+54.1 rating
- AA IntelligenceDeepSeek-V3 by 3.79.76+3.7
- vectara_answer_rateMinistral 3 14B by 2.197.599.6+2.1
Questions people ask
Which is better, DeepSeek-V3 or Ministral 3 14B?
DeepSeek-V3 wins five of the six areas where both have results: coding, reasoning, facts, long documents and following instructions. Ministral 3 14B wins agents, and costs 65% less.
Which is better for coding?
DeepSeek-V3. It wins 3 of the 3 coding tests both models report; Ministral 3 14B wins none.
Which is cheaper?
DeepSeek-V3 costs $0.24 per million input tokens and $0.90 per million output tokens; Ministral 3 14B costs $0.20 and $0.20. That makes Ministral 3 14B about 65% cheaper for the same work.
How do you compare the two?
We use the 26 benchmark tests both models have published scores on. The verdict counts the 14 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 12 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.