AI model benchmarks

Compare AI models through Arena preference charts, capability benchmarks, cost trade-offs, rankings and original evaluation evidence.

Reports, evaluations & safety observations

Report statistics, measured evaluations and reported incidents retain their own meaning.

AI Incident Database · 1,801

Stanford AI Index · 2

AILuminate · 32

Hugging Face Evaluations · 2,078

LiveCodeBench · 35

τ-bench · 85

Terminal-Bench · 61

ClockQA · exact match

AI Index 2026 · 2. Technical Performance · exact match · percent

Saxena et al., 2025

  • Qwen2-VL-7B0% 0%
  • Llama 3.2-Vision3.23% 3.23%
  • MiniCPM-V-2.63.23% 3.23%
  • GPT-o14.84% 4.84%
  • Claude-3.5-sonnet6.45% 6.45%
  • GPT-4o8.06% 8.06%
  • Gemini-2.022.58% 22.58%
Official report and methodology ↗

Technical performance relative to human baselines

AI Index 2026 · 2. Technical Performance · performance relative to human baseline · relative to human baseline

AI Index, 2026

20122025Human baseline = 1AIMEGPQA DiamondImageNet Top-5MATHMMLUMMMUOSWorldSQuAD 2.0SWE-bench VerifiedSuperGLUEVQA
  • AIME
  • GPQA Diamond
  • ImageNet Top-5
  • MATH
  • MMLU
  • MMMU
  • OSWorld
  • SQuAD 2.0
  • SWE-bench Verified
  • SuperGLUE
  • VQA
Official report and methodology ↗

Average score

bigCode · BigCode Models Leaderboard · Average score · percent · Higher is better

  1. codellama/CodeLlama-70b-Instruct-hfcodellama42.64%
  2. codellama/CodeLlama-70b-Python-hfcodellama39.61%
  3. tiiuae/falcon-180Btiiuae24.08%
  4. microsoft/phi-1microsoft12.15%
  5. Deci/DeciCoder-1bdeci5.86%
  6. bigcode/santacoderbigcode4.92%

java

bigCode · BigCode Models Leaderboard · java · percent · Higher is better

  1. Qwen2.5 Coder 32B Instructqwen73.69%
  2. Qwen/Qwen2.5-Coder-32Bqwen65.49%

lua

bigCode · BigCode Models Leaderboard · lua · percent · Higher is better

  1. Qwen/CodeQwen1.5-7Bqwen46.29%
  2. NTQAI/Nxcode-CQ-7B-orpontqai37.24%
  3. Qwen/CodeQwen1.5-7B-Chatqwen32.89%

Win Rate

bigCode · BigCode Models Leaderboard · Win Rate · percent · Higher is better

  1. NTQAI/Nxcode-CQ-7B-orpontqai55.92%
  2. Qwen/CodeQwen1.5-7B-Chatqwen55.67%
  3. Qwen/CodeQwen1.5-7Bqwen45.17%

julia

bigCode · BigCode Models Leaderboard · julia · percent · Higher is better

  1. Phind/Phind-CodeLlama-34B-v2phind45.12%
  2. Phind/Phind-CodeLlama-34B-Python-v1phind44.23%
  3. Phind/Phind-CodeLlama-34B-v1phind43.23%
  4. deepseek-ai/deepseek-coder-33b-instructdeepseek42.92%
  5. codellama/CodeLlama-70b-hfcodellama42.24%
  6. WizardLM/WizardCoder-Python-34B-V1.0wizardlm41.51%
  7. deepseek-ai/deepseek-coder-6.7b-instructdeepseek38.84%
  8. bigcode/starcoder2-15bbigcode33.19%
  9. deepseek-ai/deepseek-coder-33b-basedeepseek32.83%
  10. codellama/CodeLlama-34b-Instruct-hfcodellama32.65%
  11. codellama/CodeLlama-13b-Instruct-hfcodellama32.23%
  12. codellama/CodeLlama-34b-hfcodellama31.4%

humaneval-python

bigCode · BigCode Models Leaderboard · humaneval-python · percent · Higher is better

  1. NTQAI/Nxcode-CQ-7B-orpontqai87.23%
  2. Qwen/CodeQwen1.5-7B-Chatqwen87.2%
  3. Qwen/CodeQwen1.5-7Bqwen50.79%

lua

bigCode · BigCode Models Leaderboard · lua · percent · Higher is better

  1. codellama/CodeLlama-70b-Python-hfcodellama44.72%
  2. codellama/CodeLlama-70b-Instruct-hfcodellama44.1%
  3. tiiuae/falcon-180Btiiuae26.71%
  4. microsoft/phi-1microsoft6.21%
  5. bigcode/santacoderbigcode0.1%
  6. Deci/DeciCoder-1bdeci0%

rust

bigCode · BigCode Models Leaderboard · rust · percent · Higher is better

  1. codellama/CodeLlama-70b-Python-hfcodellama48.45%
  2. codellama/CodeLlama-70b-Instruct-hfcodellama47.2%
  3. tiiuae/falcon-180Btiiuae25%
  4. microsoft/phi-1microsoft4.49%
  5. bigcode/santacoderbigcode2%
  6. Deci/DeciCoder-1bdeci1.72%

javascript

bigCode · BigCode Models Leaderboard · javascript · percent · Higher is better

  1. Phind/Phind-CodeLlama-34B-Python-v1phind66.24%
  2. deepseek-ai/deepseek-coder-6.7b-instructdeepseek65.8%
  3. Phind/Phind-CodeLlama-34B-v2phind65.34%
  4. deepseek-ai/deepseek-coder-33b-instructdeepseek65.13%
  5. Phind/Phind-CodeLlama-34B-v1phind64.45%
  6. codellama/CodeLlama-70b-hfcodellama56.52%
  7. WizardLM/WizardCoder-Python-34B-V1.0wizardlm55.28%
  8. deepseek-ai/deepseek-coder-33b-basedeepseek51.28%
  9. deepseek-ai/deepseek-coder-6.7b-basedeepseek45.9%
  10. codellama/CodeLlama-34b-Instruct-hfcodellama45.85%
  11. codellama/CodeLlama-34b-Python-hfcodellama44.72%
  12. bigcode/starcoder2-15bbigcode44.24%

d

bigCode · BigCode Models Leaderboard · d · percent · Higher is better

  1. codellama/CodeLlama-70b-Python-hfcodellama21.12%
  2. codellama/CodeLlama-70b-Instruct-hfcodellama19.88%
  3. tiiuae/falcon-180Btiiuae14.1%
  4. microsoft/phi-1microsoft7.05%
  5. Deci/DeciCoder-1bdeci6.08%
  6. bigcode/santacoderbigcode0%

Throughput (tokens/s) bs=50

bigCode · BigCode Models Leaderboard · Throughput (tokens/s) bs=50 · tokensPerSecond · Higher is better

  1. bigcode/starcoderbase-1bbigcode2,360
  2. bigcode/starcoderbase-3bbigcode1,770
  3. bigcode/starcoderbase-7bbigcode1,700
  4. bigcode/octocoderbigcode1,520
  5. bigcode/starcoderbigcode1,490
  6. WizardLM/WizardCoder-15B-V1.0wizardlm1,470
  7. bigcode/starcoderbasebigcode1,460

Win Rate

bigCode · BigCode Models Leaderboard · Win Rate · percent · Higher is better

  1. codellama/CodeLlama-70b-Instruct-hfcodellama43.58%
  2. codellama/CodeLlama-70b-Python-hfcodellama42.33%
  3. tiiuae/falcon-180Btiiuae24.2%
  4. Deci/DeciCoder-1bdeci7.65%
  5. microsoft/phi-1microsoft6.92%
  6. bigcode/santacoderbigcode6.27%

Models

Loading
EVIDENCE / COMPARISON

Benchmarks

Explore every benchmark

Loading

These charts contain model-linked results. The charts above also include unlinked source profiles. Compare scores within each evaluation. Model quality, usage, transparency and system performance measure different things.

Install TensorTrace