# Google Gemma 3 27B As of October 21, 2026, Google Gemma 3 27B ranks #110 of 123 models overall with 31.4 ±16.1 on the LLMBenchmark Index. Gemma 3 27B ranks #103 of 122 in reasoning, its strongest category. Google Gemma 3 27B costs $0.08 per million input tokens and $0.30 per million output tokens. Gemma 3 27B was released on March 12, 2025. Page: https://llmbenchmark.ai/models/gemma-3-27b-it Data from October 21, 2026. Every figure below names the board it was read from and the day. - Maker: Google - Released: March 12, 2025 - Price per million input tokens: $0.08 - Price per million output tokens: $0.30 - Context window: 131K tokens - Overall rank: 110 of 123, by the LLMBenchmark Index - LLMBenchmark Index: 31.4 ±16.1 (90% margin, from 4 of 7 inputs) - Artificial Analysis Intelligence Index: 4.9 Benchmark results (66 of 195): - Aider polyglot: 4.9% (source: Aider polyglot leaderboard, read October 21, 2026) - AA Intelligence Index: 4.9 (source: Artificial Analysis (via OpenRouter), read October 21, 2026) - AA Coding Index: 10.1 (source: Artificial Analysis (via OpenRouter), read October 21, 2026) - AA Agentic Index: 0.1 (source: Artificial Analysis (via OpenRouter), read October 21, 2026) - AA Math Index: 20.7 (source: Artificial Analysis, read October 21, 2026) - GPQA Diamond (Artificial Analysis run): 42.8% (source: Artificial Analysis, read October 21, 2026) - Humanity's Last Exam (Artificial Analysis run): 4.4% (source: Artificial Analysis, read October 21, 2026) - MMLU Pro (Artificial Analysis run): 66.9% (source: Artificial Analysis, read October 21, 2026) - LiveCodeBench (Artificial Analysis run): 13.7% (source: Artificial Analysis, read October 21, 2026) - SciCode: 23.3% (source: Artificial Analysis, read October 21, 2026) - Terminal-Bench Hard (Artificial Analysis run): 3.8% (source: Artificial Analysis, read October 21, 2026) - IFBench: 31.8% (source: Artificial Analysis, read October 21, 2026) - Long context reasoning (AA-LCR): 7.3% (source: Artificial Analysis, read October 21, 2026) - τ²-Bench Telecom: 10.5% (source: Artificial Analysis, read October 21, 2026) - AIME 2025 (Artificial Analysis run): 20.7% (source: Artificial Analysis, read October 21, 2026) - GPQA Diamond (Epoch AI run): 47.7% (source: Epoch AI benchmarking hub, read October 21, 2026) - OTIS Mock AIME: 22.5% (source: Epoch AI benchmarking hub, read October 21, 2026) - MATH level 5: 74.0% (source: Epoch AI benchmarking hub, read October 21, 2026) - GeoBench: 52.0% (source: Epoch AI benchmarking hub, read October 21, 2026) - DTBench: 52.5% (source: Epoch AI benchmarking hub, read October 21, 2026) - LMCA: 12.3% (source: Epoch AI benchmarking hub, read October 21, 2026) - Fiction.LiveBench: 33.3% (source: Epoch AI benchmarking hub, read October 21, 2026) - Creative writing (Lech Mazur): 79.9% (source: Epoch AI benchmarking hub, read October 21, 2026) - Epoch Capabilities Index: 130.0 (source: Epoch AI benchmarking hub, read October 21, 2026) - LMArena Text: 1358 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: coding: 1322 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: hard prompts: 1340 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: math: 1312 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: instruction following: 1321 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: creative writing: 1345 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: 1165 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: English: 1377 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: non-English: 1334 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: Chinese: 1346 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: German: 1362 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: Japanese: 1287 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: Korean: 1308 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: French: 1368 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: Spanish: 1349 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: Russian: 1349 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: Polish: 1345 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: multi-turn: 1345 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: longer queries: 1333 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: expert: 1304 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: hard prompts in English: 1353 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: software and IT: 1354 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: mathematical work: 1327 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: science: 1383 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: medicine and healthcare: 1391 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: legal and government: 1375 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: business and finance: 1360 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: writing and language: 1347 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: entertainment and media: 1320 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: English: 1188 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: Chinese: 1164 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: reading text: 1156 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: diagrams: 1156 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: homework: 1160 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: humor: 1166 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: captioning: 1151 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: recognizing things: 1196 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: creative writing: 1177 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - Berkeley Function Calling Leaderboard (BFCL): 29.5% (source: Berkeley Function Calling Leaderboard, read October 21, 2026) - Kagi LLM Benchmark: 35.1% (source: Kagi LLM Benchmarking Project, read October 21, 2026) - SciCode: 0.0% (source: Kaggle Benchmarks, read October 21, 2026) - FACTS Benchmark Suite: 29.1% (source: Kaggle Benchmarks, read October 21, 2026)