Devstral Small 2 vs Gemini 3 Flash Preview
Wins 0 of 7 areas
—
Wins 6 of 7 areas
Coding · Agents · Reasoning · Images and charts · Long documents · Following instructions
Gemini 3 Flash Preview is the stronger all-rounder.
Scores updated · 20 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 software04Gemini 3 Flash Preview4 of 4 tests
- ReasoningHard problems that need careful thinking03Gemini 3 Flash Preview3 of 3 tests
- AgentsCarrying out multi-step tasks on its own02Gemini 3 Flash Preview2 of 2 tests
- Images and chartsUnderstanding pictures, charts and video01Gemini 3 Flash Preview1 of 1 test
- Long documentsFinding answers in very long texts01Gemini 3 Flash Preview1 of 1 test
- Following instructionsDoing exactly what it is asked01Gemini 3 Flash Preview1 of 1 test
- FactsGetting facts right instead of making them up11Even1 each
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.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Gemini 3 Flash Preview pulls ahead
- Reasons across sets of long documentsAA-LCR+50points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+46.8points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+37.6points ahead
Where Devstral Small 2 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+5.7points ahead
Every test, side by side
All 20 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGemini 3 Flash Preview
- Terminal-Bench 2.1Gemini 3 Flash Preview by 28.429.658+28.4
- Terminal-Bench HardGemini 3 Flash Preview by 21.916.738.6+21.9
- SciCodeGemini 3 Flash Preview by 18.232.450.6+18.2
- SWE-bench VerifiedGemini 3 Flash Preview by 106878+10
AgentsGemini 3 Flash Preview
- GDPValGemini 3 Flash Preview by 33.61.635.2+33.6
- τ-Bench V3 · BankingGemini 3 Flash Preview by 10.110.720.8+10.1
ReasoningGemini 3 Flash Preview
- GPQA DiamondGemini 3 Flash Preview by 36.653.289.8+36.6
- Humanity's Last ExamGemini 3 Flash Preview by 33.13.536.6+33.1
- CritPtGemini 3 Flash Preview by 8.608.6+8.6
FactsEven
- AA-Omniscience · AccuracyGemini 3 Flash Preview by 37.615.853.4+37.6
- AA-Omniscience · Non-hallucinationDevstral Small 2 by 5.712.77+5.7
Images and chartsGemini 3 Flash Preview
- MMMU-ProGemini 3 Flash Preview by 35.344.679.9+35.3
Long documentsGemini 3 Flash Preview
- AA-LCRGemini 3 Flash Preview by 502878+50
Following instructionsGemini 3 Flash Preview
- IFBenchGemini 3 Flash Preview by 46.831.278+46.8
Other results6 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.
- AA-OmniscienceGemini 3 Flash Preview by 67.8-57.710.1+67.8
- τ²-Bench Telecom (AA run)Gemini 3 Flash Preview by 5723.480.4+57
- AA Agentic IndexGemini 3 Flash Preview by 44.94.849.7+44.9
- Terminal-Bench 2.0Gemini 3 Flash Preview by 25.122.547.6+25.1
- AA IntelligenceGemini 3 Flash Preview by 18.87.526.3+18.8
- Artificial Analysis Coding IndexGemini 3 Flash Preview by 13.329.342.6+13.3
Questions people ask
Which is better, Devstral Small 2 or Gemini 3 Flash Preview?
Gemini 3 Flash Preview wins six of the seven areas where both have results: coding, agents, reasoning, images and charts, long documents and following instructions. Devstral Small 2 wins none. They are level on facts.
Which is better for coding?
Gemini 3 Flash Preview. It wins 4 of the 4 coding tests both models report; Devstral Small 2 wins none.
How do you compare the two?
We use the 20 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 6 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.