Gemma 4 26B A4B

Gemma 4 26B A4B is a reasoning model from Google in the Gemma 4 family. 19 benchmarks count toward its score, in 8 categories.

availableShows if the model has enough results for an index.
IndexOverall score out of 100.45.2 ±3.7
CoverageShare of the index weight with results.100%
SpeedOutput tokens per second.
Input / 1MUS dollars per 1M input tokens.Free
Output / 1MUS dollars per 1M output tokens.Free
ContextMaximum tokens in one request.256K
EloLMArena rating and rank.1434 (#80)

The index is a score out of 100. The ± range shows how much it can change. A free tier is available.

5,804 votes. Elo shows what people prefer. It does not change the score.

CapabilitiesScore per category, out of 100.

Out of 100
AgenticMulti-step tasks with tools.
33.8
CodingCode writing and repair.
47.5
ReasoningLogic problems and puzzles.
41.1
MultimodalTasks with images and text.
49.6
KnowledgeFacts and expert knowledge.
44.6
MultilingualTasks in many languages.
61.3
InstructionTasks with strict rules in the prompt.
62.6
MathMath problems.
57.5

Results

19 counted
BenchmarkThe test name.CategoryThe capability that the test measures.ResultThe score from the publisher.IndexThis result as a score out of 100.RunThe settings of the run.DateDate of the result.Published byThe source of the result.
Massive Multitask Language Understanding ProfessionalKnowledge82.6%50.7Yubo Wang et al.
OTIS Mock AIME 2024-2025Math82.2%57.5minimal effortEpoch AI
Artificial Analysis GPQA DiamondKnowledge79.2%49.7Artificial Analysis
Massive Multi-discipline Multimodal Understanding ProMultimodal73.8%47.4MMMU-Pro authors
GPQA diamondKnowledge73.2%45.9minimal effortEpoch AI
Artificial Analysis IFBenchInstruction72.4%62.6Artificial Analysis
Artificial Analysis MMMU-ProMultimodal69.2%51.7Artificial Analysis
Artificial Analysis Long Context ReasoningReasoning65.7%53.7Artificial Analysis
EuroEval FrenchMultilingual65.2%83.7EuroEval
EuroEval PortugueseMultilingual62.0%79.6EuroEval
EuroEval ItalianMultilingual61.1%78.6EuroEval
EuroEval SwedishMultilingual58.8%75.7EuroEval
EuroEval DutchMultilingual56.4%72.7EuroEval
EuroEval PolishMultilingual55.5%71.6EuroEval
EuroEval SpanishMultilingual49.4%64.0EuroEval
EuroEval GermanMultilingual49.1%63.7EuroEval
EuroEval DutchMultilingual45.3%58.9EuroEval
EuroEval FrenchMultilingual44.4%57.8EuroEval
EuroEval SwedishMultilingual44.4%57.8EuroEval
τ²-Bench Tool-Agent-User EvaluationAgentic43.6%30.1Victor Barres et al.
EuroEval PortugueseMultilingual43.5%56.7EuroEval
EuroEval ItalianMultilingual41.0%53.5EuroEval
Artificial Analysis SciCodeCoding40.0%48.4Artificial Analysis
Artificial Analysis Coding IndexCoding39.3%46.7Artificial Analysis
EuroEval PolishMultilingual38.5%50.4EuroEval
EuroEval GermanMultilingual37.7%49.4EuroEval
EuroEval SpanishMultilingual35.4%46.5EuroEval
Artificial Analysis Humanity's Last ExamKnowledge19.3%45.5Artificial Analysis
Artificial Analysis Omniscience AccuracyKnowledge19.1%37.5Artificial Analysis
Humanity's Last ExamKnowledge17.2%43.4Center for AI Safety et al.
Artificial Analysis Intelligence IndexKnowledge16.7%43.2Artificial Analysis
Humanity's Last Exam without toolsKnowledge8.7%36.2OpenAI
Chess PuzzlesReasoning6.0%31.3minimal effortEpoch AI
GDPval-AA normalizedAgentic2.6%37.6Artificial Analysis
Critical Physics TasksReasoning0.0%38.4Artificial Analysis

19 benchmarks count, from 35 of 35 results. A grey row does not count. Too few models took that benchmark.

Sources

BenchLM benchmark aggregationUsed with attribution; per-benchmark results credited to their original authorsEpoch AI, collected directlyCC BY — free to use and redistribute with attributionEuroEval, collected directlyMIT — the leaderboard site and its CSV routes are in the licensed repository

More from Google

Gemini 3.7 Flash72.3Gemini 3.8 Flash72.1Gemini 3.5 Flash67.3Gemini 3.6 Flash66.8Gemini 3.1 Pro64.7