Gemma 3 12B vs Mistral Large 3
Wins 0 of 7 areas
—
Wins 6 of 7 areas
Coding · Agents · Reasoning · Facts · Images and charts · Long documents
Mistral Large 3 is the stronger all-rounder.Gemma 3 12B is cheaper.
Scores updated · 27 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 software03Mistral Large 33 of 3 tests
- AgentsCarrying out multi-step tasks on its own01Mistral Large 31 of 3 tests · 2 ties
- ReasoningHard problems that need careful thinking01Mistral Large 31 of 3 tests · 2 ties
- FactsGetting facts right instead of making them up12Mistral Large 32 of 3 tests
- Images and chartsUnderstanding pictures, charts and video01Mistral Large 31 of 1 test
- Long documentsFinding answers in very long texts01Mistral Large 31 of 1 test
- Following instructionsDoing exactly what it is asked00Even0 each · 1 tie
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 12B costs 90% 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 Mistral Large 3 pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+33.1points ahead
- Reasons across sets of long documentsAA-LCR+27.7points ahead
- Code for real scientific research problemsSciCode+20.2points ahead
Where Gemma 3 12B pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 27 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingMistral Large 3
- SciCodeMistral Large 3 by 20.216.436.6+20.2
- Terminal-Bench HardMistral Large 3 by 15.10.815.9+15.1
- Terminal-Bench 2.1Mistral Large 3 by 12012+12
AgentsMistral Large 3
- τ-Bench V3 · BankingMistral Large 3 by 50.85.8+5
- GDPValtie00tie
- Terminal-Bench 4.0tie00tie
ReasoningMistral Large 3
- GPQA DiamondMistral Large 3 by 33.134.968+33.1
- CritPttie00tie
- Humanity's Last Examtie4.24.2tie
FactsMistral Large 3
- AA-Omniscience · AccuracyMistral Large 3 by 14.710.324.9+14.7
- AA-Omniscience · Non-hallucinationMistral Large 3 by 11.42.614+11.4
- Vectara HHEM hallucination ratelower is betterGemma 3 12B by 10.14.414.5+10.1
Images and chartsMistral Large 3
- MMMU-ProMistral Large 3 by 18.237.555.7+18.2
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.
- MM MTBenchMistral Large 3 by 78.26.784.9+78.2
- AA-OmniscienceMistral Large 3 by 37.5-77.1-39.6+37.5
- SimpleQAMistral Large 3 by 17.56.323.8+17.5
- Artificial Analysis Coding IndexMistral Large 3 by 14.35.820.1+14.3
- τ²-Bench Telecom (AA run)Mistral Large 3 by 13.810.824.6+13.8
- Arena HardMistral Large 3 by 11.543.655.1+11.5
- vectara_factual_consistencyGemma 3 12B by 10.195.685.5+10.1
- LiveCodeBenchMistral Large 3 by 9.824.634.4+9.8
- AA IntelligenceMistral Large 3 by 5.53.89.3+5.5
- WildBenchMistral Large 3 by 5.363.268.5+5.3
- vectara_avg_summary_lengthMistral Large 3 by 23 rating points89.7112.7+23 rating
- vectara_answer_rateMistral Large 3 by 1.497.498.8+1.4
Questions people ask
Which is better, Gemma 3 12B or Mistral Large 3?
Mistral Large 3 wins six of the seven areas where both have results: coding, agents, reasoning, facts, images and charts and long documents. Gemma 3 12B wins none, but costs 90% less. They are level on following instructions.
Which is better for coding?
Mistral Large 3. It wins 3 of the 3 coding tests both models report; Gemma 3 12B wins none.
Which is cheaper?
Gemma 3 12B costs $0.05 per million input tokens and $0.15 per million output tokens; Mistral Large 3 costs $0.50 and $1.50. That makes Gemma 3 12B about 90% cheaper for the same work.
How do you compare the two?
We use the 27 benchmark tests both models have published scores on. The verdict counts the 15 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.