DeepSeek quietly ships V4 Pro GA, claims near-parity with Claude Fable 5
DeepSeek released the finished DeepSeek-V4-Pro-0813 build on Wednesday with no announcement, replacing the April preview. Its benchmark table puts Anthropic's Claude Fable 5 ahead by an average of 5.3% across nine agent benchmarks, while Fable 5 costs $10/$50 per million input/output tokens versus V4 Pro's $0.435/$0.87.
1 source
Daily brief
Get tomorrow's AI brief in your inbox
More stories today
- Robotic systems struggle with physical navigation in recent demonstrations
- ChatGPT Visualize skill turns notes into interactive interfaces
- Twitch streamers sue Twitch, Amazon over AI training data use
- Fastino launches GLiNER2.5 with boundary-prediction architecture
- NVIDIA Sol Engine cuts MiniMax H3 latency to 14.93s