Sahara logo





Main Leaderboard


model Sahara Score Text Classification Tasks Text Generation Tasks MCCR Tasks Tokens Level Tasks
🏆 gemini-3-pro-preview 55.67 52.67 18.24 78.58 73.18
🥈 gemini-2.5-pro 51.85 48.18 15.66 77.56 65.99
🥉 gpt-5-2025-08-07 50.03 48.19 11.10 75.77 65.07
gpt-5.1-2025-11-13 48.33 47.77 10.32 73.64 61.59
GPT-5.6 Luna 47.53 46.45 13.54 68.50 61.65
claude-4-sonnet-20250514 40.82 47.28 10.59 60.53 44.86
gemini-2.5-flash 40.61 43.95 11.18 72.41 34.91
claude-sonnet-4-5@20250929 37.11 45.15 12.04 52.41 38.85
gpt-4.1 36.04 48.07 11.06 50.98 34.05
Qwen/Qwen3.5-27B 35.96 43.64 11.73 54.61 33.84
Qwen/Qwen3.6-27B 35.41 43.20 11.96 54.06 32.43
Qwen/Qwen3.6-35B-A3B 32.09 40.43 12.42 51.55 23.95
CohereForAI/c4ai-command-a-03-2025 29.93 38.64 10.36 45.55 25.16
meta-llama/Llama-4-Scout-17B-16E-Instruct 29.19 39.02 11.45 47.36 18.94
google/gemma-3-27b-it 28.07 44.44 8.19 43.20 16.45
meta-llama/Llama-3.3-70B-Instruct 27.05 37.41 9.80 44.77 16.24
Qwen/Qwen3.5-9B 26.61 37.72 10.54 41.48 16.71
meta-llama/Llama-3.1-70B-Instruct 26.57 35.96 11.15 43.66 15.51
Google/gemma-4-31B-it 26.18 23.49 10.15 42.74 28.33
Qwen/Qwen3.5-4B 24.79 33.88 11.08 37.72 16.46
Qwen/Qwen3-30B-Instruct-2507 24.73 30.01 9.79 42.28 16.85
deepseek-ai/DeepSeek-R1-Distill-Llama-70B 24.42 34.96 10.91 39.50 12.31
deepseek-ai/DeepSeek-R1-Distill-Qwen-32B 24.08 29.27 9.61 39.70 17.74
google/gemma-2-27b-it 23.80 24.53 9.01 40.93 20.72
CohereForAI/c4ai-command-r-plus-08-2024 23.21 33.72 10.85 32.45 15.82
google/gemma-2-9b-it 22.80 28.10 8.54 37.88 16.68
google/gemma-3-12b-it 22.25 33.55 8.36 35.95 11.16
McGill-NLP/AfriqueQwen-14B 22.13 24.85 10.78 39.88 13.04
Qwen/Qwen3-4B-Instruct-2507 21.49 28.88 9.93 33.74 13.41
Tower-Babel/Babel-83B-Chat 21.25 22.89 7.52 37.27 17.31
meta-llama/Llama-3.1-8B-Instruct 21.08 31.41 9.92 33.88 9.11
McGill-NLP/AfriqueQwen3.5-4B 20.57 28.38 10.37 32.52 11.02
Tower-Babel/Babel-9B-Chat 19.61 24.63 9.35 33.73 10.73
Qwen/Qwen3.5-2B 19.44 26.57 9.58 30.23 11.39
McGill-NLP/AfriqueQwen-8B 19.41 22.57 6.65 38.02 10.39
CohereForAI/c4ai-command-r7b-12-2024 18.06 23.53 5.84 28.34 14.53
CohereForAI/aya-23-35B 17.75 22.36 7.06 29.56 12.00
Google/gemma-4-E4B-it 17.38 17.17 7.95 32.45 11.94
CohereForAI/aya-23-8B 17.30 21.94 6.55 26.86 13.86
google/gemma-2-2b-it 17.16 17.69 6.73 30.30 13.92
google/gemma-3-4b-it 17.10 14.65 6.98 32.84 13.95
microsoft/Phi-4-mini-instruct 16.78 16.50 5.10 33.73 11.78
meta-llama/Llama-3.2-3B-Instruct 16.49 18.08 6.56 29.33 12.00
google_gemma_4_26B_A4B_it 14.77 5.67 4.95 27.55 20.89
meta-llama/Llama-3.2-1B-Instruct 14.64 16.34 6.39 27.96 7.87
microsoft/Phi-3.5-mini-instruct 14.19 17.96 6.03 25.23 7.52
Google/gemma-4-E2B-it 12.44 10.45 8.98 24.38 5.95

Citation

If you use the Sahara benchmark for your scientific publication, or if you find the resources in this website useful, please cite our ACL2025 paper .