HEAD-TO-HEAD MODEL ARENA

Llama 3.3 70B Instruct vs Qwen 2.5 72B Instruct

Side-by-side benchmark differential across coding SWE-bench, human battle Arena ELO, throughput, and API token spend.

Llama 3.3 70B Instruct

Meta

  • 🏆 Arena Elo: 1285 pts
  • 💻 SWE-bench: 82/100
  • 🤖 Agentic: 84/100
  • 💵 Input: $0.12/1M
  • 💵 Output: $0.30/1M

Qwen 2.5 72B Instruct

Alibaba Qwen

  • 🏆 Arena Elo: 1290 pts
  • 💻 SWE-bench: 86.5/100
  • 🤖 Agentic: 85/100
  • 💵 Input: $0.35/1M
  • 💵 Output: $0.40/1M