Gemma 3 4B vs Ministral 3 14B
Wins 0 of 7 areas
—
Wins 7 of 7 areas
Coding · Agents · Reasoning · Facts · Images and charts · Long documents · Following instructions
Ministral 3 14B is the stronger all-rounder.Gemma 3 4B is cheaper.
Scores updated · 25 tests both models report · How we compare
Where each one wins
Tests won in each of the seven 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 software02Ministral 3 14B2 of 2 tests
- ReasoningHard problems that need careful thinking01Ministral 3 14B1 of 3 tests · 2 ties
- FactsGetting facts right instead of making them up12Ministral 3 14B2 of 3 tests
- AgentsCarrying out multi-step tasks on its own01Ministral 3 14B1 of 2 tests · 1 tie
- Images and chartsUnderstanding pictures, charts and video01Ministral 3 14B1 of 1 test
- Long documentsFinding answers in very long texts01Ministral 3 14B1 of 1 test
- Following instructionsDoing exactly what it is asked01Ministral 3 14B1 of 1 test
Images and charts, 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.
Gemma 3 4B costs 63% 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 Ministral 3 14B pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+28.1points ahead
- Harder college exam questions with imagesMMMU-Pro+19.9points ahead
- Reasons across sets of long documentsAA-LCR+19.6points ahead
Where Gemma 3 4B pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 25 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingMinistral 3 14B
- SciCodeMinistral 3 14B by 16.57.323.8+16.5
- Terminal-Bench HardMinistral 3 14B by 3.70.84.5+3.7
AgentsMinistral 3 14B
- τ-Bench V3 · BankingMinistral 3 14B by 6.20.46.6+6.2
- GDPValtie00tie
ReasoningMinistral 3 14B
- GPQA DiamondMinistral 3 14B by 28.129.157.2+28.1
- Humanity's Last Examtie5.34.6tie
- CritPttie00tie
FactsMinistral 3 14B
- Vectara HHEM hallucination ratelower is betterGemma 3 4B by 136.419.4+13
- AA-Omniscience · AccuracyMinistral 3 14B by 5.97.713.6+5.9
- AA-Omniscience · Non-hallucinationMinistral 3 14B by 5.71.87.5+5.7
Images and chartsMinistral 3 14B
- MMMU-ProMinistral 3 14B by 19.929.949.8+19.9
Following instructionsMinistral 3 14B
- IFBenchMinistral 3 14B by 3.728.332+3.7
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.
- LiveCodeBenchMinistral 3 14B by 5212.664.6+52
- vectara_answer_rateMinistral 3 14B by 32.367.399.6+32.3
- Arena HardMinistral 3 14B by 23.331.855.1+23.3
- τ²-Bench Telecom (AA run)Ministral 3 14B by 22.2527.2+22.2
- MMLUMinistral 3 14B by 2158.479.4+21
- WildBenchMinistral 3 14B by 19.449.168.5+19.4
- AA-OmniscienceMinistral 3 14B by 16.5-82.9-66.4+16.5
- vectara_factual_consistencyGemma 3 4B by 1393.680.6+13
- Artificial Analysis Coding IndexMinistral 3 14B by 11.72.714.4+11.7
- vectara_avg_summary_lengthMinistral 3 14B by 58.4 rating points77.4135.8+58.4 rating
- MM MTBenchMinistral 3 14B by 3.35.28.5+3.3
- AA IntelligenceMinistral 3 14B by 1.24.86+1.2
Questions people ask
Which is better, Gemma 3 4B or Ministral 3 14B?
Ministral 3 14B wins all seven areas where both have results: coding, agents, reasoning, facts, images and charts, long documents and following instructions. Gemma 3 4B wins none, but costs 63% less.
Which is better for coding?
Ministral 3 14B. It wins 2 of the 2 coding tests both models report; Gemma 3 4B wins none.
Which is cheaper?
Gemma 3 4B costs $0.05 per million input tokens and $0.10 per million output tokens; Ministral 3 14B costs $0.20 and $0.20. That makes Gemma 3 4B about 63% cheaper for the same work.
How do you compare the two?
We use the 25 benchmark tests both models have published scores on. The verdict counts the 13 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.