DeepSeek-V3 vs Gemma 4 26B A4B
Wins 1 of 6 areas
Facts
Wins 5 of 6 areas
Coding · Agents · Reasoning · Long documents · Following instructions
Gemma 4 26B A4B is the stronger all-rounder.DeepSeek-V3 is better at facts.
Scores updated · 38 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.
- ReasoningHard problems that need careful thinking02Gemma 4 26B A4B2 of 3 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own02Gemma 4 26B A4B2 of 2 tests
- CodingWriting and fixing software23Gemma 4 26B A4B3 of 5 tests
- Long documentsFinding answers in very long texts01Gemma 4 26B A4B1 of 1 test
- Following instructionsDoing exactly what it is asked01Gemma 4 26B A4B1 of 1 test
- FactsGetting facts right instead of making them up10DeepSeek-V31 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.
Gemma 4 26B A4B costs 54% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Gemma 4 26B A4B pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+31.4points ahead
- Recent programming contest problemsLiveCodeBench v6+30.2points ahead
- Reasons across sets of long documentsAA-LCR+25points ahead
Where DeepSeek-V3 pulls ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+24.6points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+6.4points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+1.6points ahead
Every test, side by side
All 38 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGemma 4 26B A4B
- LiveCodeBench v6Gemma 4 26B A4B by 30.246.977.1+30.2
- SWE-bench VerifiedDeepSeek-V3 by 24.64217.4+24.6
- Terminal-Bench 2.1Gemma 4 26B A4B by 22.116.939+22.1
- Terminal-Bench HardDeepSeek-V3 by 1.615.213.6+1.6
- SciCodeGemma 4 26B A4B by 13940+1
AgentsGemma 4 26B A4B
- τ-Bench V3 · BankingGemma 4 26B A4B by 7.34.712+7.3
- GDPValGemma 4 26B A4B by 3.403.4+3.4
ReasoningGemma 4 26B A4B
- Humanity's Last ExamGemma 4 26B A4B by 14.64.719.3+14.6
- GPQA DiamondGemma 4 26B A4B by 13.765.579.2+13.7
- CritPttie00tie
FactsDeepSeek-V3
- AA-Omniscience · AccuracyDeepSeek-V3 by 6.425.419.1+6.4
- Vectara HHEM hallucination ratelower is bettertie6.15.2tie
- AA-Omniscience · Non-hallucinationtie14.113.6tie
Following instructionsGemma 4 26B A4B
- IFBenchGemma 4 26B A4B by 31.44172.4+31.4
Other results23 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.
- HMMT Feb. 2025Gemma 4 26B A4B by 62.529.291.7+62.5
- OpenBookQADeepSeek-V3 by 46.695.448.8+46.6
- τ²-BenchGemma 4 26B A4B by 35.732.568.2+35.7
- Artificial Analysis Coding IndexGemma 4 26B A4B by 16.32339.3+16.3
- HumanEvalDeepSeek-V3 by 15.265.250+15.2
- vectara_avg_summary_lengthDeepSeek-V3 by 14.681.767.1+14.6
- MMLUDeepSeek-V3 by 10.788.577.8+10.7
- AA-OmniscienceDeepSeek-V3 by 10.1-40.7-50.8+10.1
- SuperGPQAGemma 4 26B A4B by 7.753.761.4+7.7
- AA IntelligenceGemma 4 26B A4B by 79.716.7+7
- MBPPDeepSeek-V3 by 775.468.4+7
- MMMLUGemma 4 26B A4B by 6.979.486.3+6.9
- MMLU-ProGemma 4 26B A4B by 6.775.982.6+6.7
- WinoGrandeDeepSeek-V3 by 5.884.979.1+5.8
- IFEvalGemma 4 26B A4B by 5.386.191.4+5.3
- C-EvalDeepSeek-V3 by 486.582.5+4
- HellaSwagDeepSeek-V3 by 3.688.985.3+3.6
- MMLU-ReduxGemma 4 26B A4B by 3.689.192.7+3.6
- τ²-Bench Telecom (AA run)DeepSeek-V3 by 3.547.143.6+3.5
- ARC-ChallengeDeepSeek-V3 by 2.595.392.8+2.5
- PIQA (Acc.)DeepSeek-V3 by 2.584.782.2+2.5
- vectara_answer_rateGemma 4 26B A4B by 2.397.599.8+2.3
- vectara_factual_consistencytie93.994.8tie
Questions people ask
Which is better, DeepSeek-V3 or Gemma 4 26B A4B?
Gemma 4 26B A4B wins five of the six areas where both have results: coding, agents, reasoning, long documents and following instructions. DeepSeek-V3 wins facts.
Which is better for coding?
Gemma 4 26B A4B. It wins 3 of the 5 coding tests both models report; DeepSeek-V3 wins 2.
Which is cheaper?
DeepSeek-V3 costs $0.24 per million input tokens and $0.90 per million output tokens; Gemma 4 26B A4B costs $0.13 and $0.40. That makes Gemma 4 26B A4B about 54% cheaper for the same work.
How do you compare the two?
We use the 38 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 23 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.