Note: Overall leaderboard rankings may not reflect true model quality — individual benchmarks give a clearer picture. ARC-Challenge MMLU GPQA GSM8K Artificial Analysis Intelligence Index v4.1
← Back to leaderboard

Google: Gemma 4 E2B Instruct

10 benchmarks
CompanyGoogle
Country of originUSA

Leaderboard position

#765
Overall rank (of 1122 models)
543
Elo rating
10
Benchmarks
Top 68.2%
Percentile
#54
In Google (of 73 models)
Above: Yandex: gpt-5-lite-pretrain #764 Below: OpenAI: gpt-4o-mini #766

Benchmark highlights

Best benchmark
MMLU-Pro 60
Worst benchmark
AA-Omniscience Index -24

Nearest competitor from another company

The model from another company with the most similar benchmark scores.

appeared as nearest in 1 of 10 benchmarks

perplexity: sonar-pro
1/10 10.0%
Technology Innovation Institute: falcon-h1r-7b
1/10 10.0%
Qwen: Qwen3 0.6B
1/10 10.0%
JetBrains: mellum2-12b-a2.5b-base
1/10 10.0%
Moonshot AI: kimi-linear-48b-a3b-instruct
1/10 10.0%
Xiaomi: mimo-v2-pro
1/10 10.0%
SpaceXAI: Grok 4.1 Fast
1/10 10.0%
state-spaces: mamba-130m
1/10 10.0%
Model Votes % of benchmarks
perplexity: sonar-pro 1 10.0%
Technology Innovation Institute: falcon-h1r-7b 1 10.0%
Qwen: Qwen3 0.6B 1 10.0%
JetBrains: mellum2-12b-a2.5b-base 1 10.0%
Moonshot AI: kimi-linear-48b-a3b-instruct 1 10.0%
Xiaomi: mimo-v2-pro 1 10.0%
SpaceXAI: Grok 4.1 Fast 1 10.0%
state-spaces: mamba-130m 1 10.0%

Benchmarks

MMLU-Pro 60 CoVoST 33.47 Artificial Analysis Intelligence Index v4.0 (Maximum Reasoning) 15.2 Artificial Analysis Coding Index (Maximum Reasoning) 9 Artificial Analysis Intelligence Index v4.1 (Maximum Reasoning) 9 Artificial Analysis Agentic Index (Maximum Reasoning) 6.9 CritPt 0 -(FLEURS) -0.09 -(CoderPPL) -3.8687 AA-Omniscience Index -24

Predicted scores

Estimated from 25 similar models and the global Elo rating. These are projections, not measured results.

Benchmark Predicted Confidence Based on
Artificial Analysis Intelligence Index v4.3.2 ~9 64% LiquidAI: lfm-2.5-8b-a1b, OpenBMB: minicpm5-1b, Qwen: Qwen3.5 2B
GSM8K ~70.78 63% IBM Granite: granite-4.1-30b, DeepSeek: deepseek-v3, IBM Granite: Granite 4.1 8B
CowQA ~58.85 63% LiquidAI: lfm-2.5-8b-a1b, IBM Granite: Granite 4.1 8B, Qwen: Qwen3 4B
MATH-500 ~81.53 62% LiquidAI: lfm-2.5-8b-a1b, Technology Innovation Institute: falcon-h1r-7b, Qwen: Qwen3 4B
AIME 2025 ~55.57 62% LiquidAI: lfm-2.5-8b-a1b, Technology Innovation Institute: falcon-h1r-7b, Qwen: Qwen3 4B
GPQA Diamond ~50.6 62% Meta Llama: Llama 3.3 70B Instruct, Technology Innovation Institute: falcon-h1r-7b, Qwen: Qwen3 4B
IFBench ~48.05 62% LiquidAI: lfm-2.5-8b-a1b, Meta Llama: Llama 3.3 70B Instruct, Technology Innovation Institute: falcon-h1r-7b
LiveCodeBench v6 ~48.75 62% Qwen: Qwen3 4B, Qwen: qwen3-1.7b, Google: Gemma 4 E4B Instruct
Humanity's Last Exam (no tools) ~6.2 61% Meta Llama: Llama 3.3 70B Instruct, Technology Innovation Institute: falcon-h1r-7b, Microsoft: phi-4-mini-instruct
IFEval ~77.67 61% LiquidAI: lfm-2.5-8b-a1b, Qwen: Qwen3 4B, OpenBMB: minicpm5-1b
MMLU ~58.62 61% LiquidAI: lfm-2.5-8b-a1b, OpenBMB: minicpm5-1b, Qwen: Qwen3.5 2B
SciCode ~22.29 61% Meta Llama: Llama 3.3 70B Instruct, Technology Innovation Institute: falcon-h1r-7b, SpaceXAI: Grok 3
ARC-Easy ~74.44 61% LiquidAI: lfm-2.5-8b-a1b, OpenBMB: minicpm5-1b, Qwen: Qwen3.5 2B
ARC-Challenge ~59.37 61% LiquidAI: lfm-2.5-8b-a1b, OpenBMB: minicpm5-1b, Qwen: Qwen3.5 2B
HellaSwag ~71.12 61% LiquidAI: lfm-2.5-8b-a1b, OpenBMB: minicpm5-1b, Qwen: Qwen3.5 2B
SuperGPQA ~33.58 61% Meta Llama: Llama 3.3 70B Instruct, DeepSeek: deepseek-v3, OpenBMB: minicpm5-1b
WikiText-2 (-ppl) ~-24.44 61% LiquidAI: lfm-2.5-8b-a1b, Mistral AI: ministral-3-3b-instruct-2512, OpenBMB: minicpm5-1b
WinoGrande ~68.96 61% LiquidAI: lfm-2.5-8b-a1b, OpenBMB: minicpm5-1b, Qwen: Qwen3.5 2B
TruthfulQA ~41 61% LiquidAI: lfm-2.5-8b-a1b, OpenBMB: minicpm5-1b, Qwen: Qwen3.5 2B
HMMT Feb 2026 ~58.98 60% Qwen: Qwen3 4B, OpenBMB: minicpm5-1b, Qwen: Qwen3.5 9B
SWE-bench Verified ~30.58 60% Qwen: Qwen3 4B, OpenAI: gpt-oss-20b, Cohere Labs: north-mini-code-1.0
AA-LCR ~36.92 60% Meta Llama: Llama 3.3 70B Instruct, SpaceXAI: Grok 3, Cohere Labs: Tiny Aya Global
GPQA ~40.07 60% OpenAI: GPT-3.5 Turbo, Qwen: Qwen3.5 0.8B, NVIDIA Corporation: nvidia-nemotron-3-nano-30b-a3b
ScreenSpot-Pro ~46.3 60% Qwen: Qwen3.5 2B, Qwen: Qwen3.5 0.8B, Qwen: Qwen3.5 4B
Text Arena (Snapshot from 20 june 2026) ~1290.86 60% Meta Llama: Llama 4 Maverick Instruct, OpenAI: GPT-3.5 Turbo, DeepSeek: DeepSeek R1