DeepSeek-V4.1-Flash Hits #3 Open Model on Agent Arena at $0.07 per Task
arena · x · 2026-09-15
Agent Arena reports that DeepSeek-V4.1-Flash (Max) landed at #3 among open models and #12 overall, with +4.87% net improvement and a median cost of $0.07 per task — reshaping the Pareto frontier for cost efficiency.
Among the top 3 open models:
- Kimi K3 (Max): +6.39% | $0.77/task
- Hy4 preview: +4.96% | $0.22/task
- DeepSeek-V4.1-Flash (Max): +4.87% | $0.07/task
It retains 98% of Hy4 preview's net improvement at 73% lower cost, and 76% of Kimi K3's performance at 92% lower cost. Against frontier models like Claude Fable 5.1 (Max) (+13.90% | $4.54) and GPT 6 Astra (Max) (+11.90% | $4.09), it keeps 35–54% of their net improvement at 97–99% lower cost — top models cost 37–76× more per task.
Related event: DeepSeek-V4.1-Flash Ranks Third Among Open Models on Agent Arena(3 posts)→
More from coding & agent
- Stop reading terminals: build with AI entirely in realtime HTML pages — nptacek · 2026-09-15
- Coding agents must survive a day without saying 'You're right, sorry' before ASI, researcher jokes — andrey_kurenkov · 2026-09-15
- Claude Code's Function Hooks bring Express-style middleware to plugins — nicolascraske · 2026-09-15
- Review agent commits in parallel with zero context via the show-me skill — lucasmeijer · 2026-09-15
- Dolphin creator Eric Hartford unveils With, a systems language for AI infrastructure — QuixiAI · 2026-09-15
- Headroom 1.4 adds motion design, new gauges, and Cursor and Grok support — kieranklaassen · 2026-09-15