• Jono Nourse
• 03 Aug, 2026
• • deepseek
  • ai-agents
  • model-selection
  • tool-calling
  • terminal-bench
  • fireworks-ai
  • cost-efficiency

DeepSeek V4-Flash Closes In on Top Agent Models at 2% of the Price

DeepSeek V4-Flash hits 82.7 on Terminal-Bench 2.1, three points behind GPT-5.6 Sol, at roughly 2% of the input cost. It's live in Autohive via …