Gemma 4 26B A4B vs Llama 3.1 Instruct 405B
Wins 5 of 6 areas
Coding · Agents · Reasoning · Long documents · Following instructions
Wins 1 of 6 areas
Facts
Gemma 4 26B A4B is the stronger all-rounder.Llama 3.1 Instruct 405B is better at facts.
Scores updated · 30 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 software31Gemma 4 26B A4B3 of 4 tests
- ReasoningHard problems that need careful thinking20Gemma 4 26B A4B2 of 3 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own10Gemma 4 26B A4B1 of 1 test
- Long documentsFinding answers in very long texts10Gemma 4 26B A4B1 of 1 test
- Following instructionsDoing exactly what it is asked10Gemma 4 26B A4B1 of 1 test
- FactsGetting facts right instead of making them up02Llama 3.1 Instruct 405B2 of 2 tests
Agents, 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 three tests each model wins by the widest margin. Scores are out of 100.
Where Gemma 4 26B A4B pulls ahead
- Reasons across sets of long documentsAA-LCR+40.4points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+33.4points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+27.7points ahead
Where Llama 3.1 Instruct 405B pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+34points ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+7.1points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+4.1points ahead
Every test, side by side
All 30 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGemma 4 26B A4B
- SciCodeGemma 4 26B A4B by 10.14029.9+10.1
- SWE-bench VerifiedLlama 3.1 Instruct 405B by 7.117.424.5+7.1
- Terminal-Bench HardGemma 4 26B A4B by 6.813.66.8+6.8
- SWE-bench ProGemma 4 26B A4B by 2.613.811.2+2.6
ReasoningGemma 4 26B A4B
- GPQA DiamondGemma 4 26B A4B by 27.779.251.5+27.7
- Humanity's Last ExamGemma 4 26B A4B by 15.319.34+15.3
- CritPttie00tie
FactsLlama 3.1 Instruct 405B
- AA-Omniscience · Non-hallucinationLlama 3.1 Instruct 405B by 3413.647.6+34
- AA-Omniscience · AccuracyLlama 3.1 Instruct 405B by 4.119.123.2+4.1
Following instructionsGemma 4 26B A4B
- IFBenchGemma 4 26B A4B by 33.472.439+33.4
Other results18 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.
- OpenBookQALlama 3.1 Instruct 405B by 45.248.894+45.2
- HumanEvalLlama 3.1 Instruct 405B by 395089+39
- AA-OmniscienceLlama 3.1 Instruct 405B by 33.7-50.8-17.1+33.7
- Artificial Analysis Coding IndexGemma 4 26B A4B by 24.839.314.5+24.8
- τ²-Bench Telecom (AA run)Gemma 4 26B A4B by 24.643.619+24.6
- MMMLUGemma 4 26B A4B by 12.586.373.8+12.5
- MMLULlama 3.1 Instruct 405B by 10.877.888.6+10.8
- C-EvalGemma 4 26B A4B by 1082.572.5+10
- AA IntelligenceGemma 4 26B A4B by 9.416.77.3+9.4
- MMLU-ProGemma 4 26B A4B by 9.382.673.3+9.3
- MMLU-ReduxGemma 4 26B A4B by 6.592.786.2+6.5
- WinoGrandeLlama 3.1 Instruct 405B by 6.179.185.2+6.1
- AA Agentic IndexGemma 4 26B A4B by 4.7116.3+4.7
- ARC-ChallengeLlama 3.1 Instruct 405B by 4.192.896.9+4.1
- HellaSwagLlama 3.1 Instruct 405B by 3.985.389.2+3.9
- PIQA (Acc.)Llama 3.1 Instruct 405B by 3.782.285.9+3.7
- IFEvalGemma 4 26B A4B by 2.891.488.6+2.8
- MBPPtie68.468.4tie
Questions people ask
Which is better, Gemma 4 26B A4B or Llama 3.1 Instruct 405B?
Gemma 4 26B A4B wins five of the six areas where both have results: coding, agents, reasoning, long documents and following instructions. Llama 3.1 Instruct 405B wins facts.
Which is better for coding?
Gemma 4 26B A4B. It wins 3 of the 4 coding tests both models report; Llama 3.1 Instruct 405B wins 1.
How do you compare the two?
We use the 30 benchmark tests both models have published scores on. The verdict counts the 12 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 18 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.