aevionai's picture
🏟️ Full Sweep LIVE: 8 models benchmarked — qwen3:8b tops at 83.3%, deepseek-coder-v2:16b ties at 83.3%
e5894f6 verified
|
raw
history blame
1.49 kB

Man in the Arena

  • Generated: 2026-04-22T14:28:59.687850+00:00
  • Git commit: 2f221ca48ab59cc48ed918255b9c739de9e99fe7
  • Chain id: arena-20260422T142859Z
Rank Model Composite Mean Latency (ms) Mean TPS
1 qwen3:8b 0.833 24997.49 55.63
2 deepseek-coder-v2:16b 0.833 49680.85 41.90
3 gemma4-26b-iq3xxs:latest 0.500 55791.52 20.46
4 qwen:latest 0.000 10547.31 109.00
5 deepseek-coder:6.7b 0.000 50345.45 71.20
6 johnnyboy/qwen2.5-math-7b:latest 0.000 10592.99 66.20
7 gemma4:26b 0.000 54956.90 13.69
8 phi4-reasoning:plus 0.000 84923.44 8.98

Suite Breakdown

qwen3:8b

  • claim_verification_f1 / claim_verification_f1: 0.833 (2 cases)

deepseek-coder-v2:16b

  • claim_verification_f1 / claim_verification_f1: 0.833 (2 cases)

gemma4-26b-iq3xxs:latest

  • claim_verification_f1 / claim_verification_f1: 0.500 (2 cases)

qwen:latest

  • claim_verification_f1 / claim_verification_f1: 0.000 (2 cases)

deepseek-coder:6.7b

  • claim_verification_f1 / claim_verification_f1: 0.000 (2 cases)

johnnyboy/qwen2.5-math-7b:latest

  • claim_verification_f1 / claim_verification_f1: 0.000 (2 cases)

gemma4:26b

  • claim_verification_f1 / claim_verification_f1: 0.000 (2 cases)

phi4-reasoning:plus

  • claim_verification_f1 / claim_verification_f1: 0.000 (2 cases)