DeepSeek V4 Flash update boosts agentic benchmarks
DeepSeek V4 Flash (2026-07-31) scores 82.7 on Terminal Bench, up from 56.9 in preview. Toolathlon rises to 70.3 from 51.8, with new benchmarks including NL2Repo (54.2) and Cybergym (76.7).
1 source
Daily brief
Get tomorrow's AI brief in your inbox
More stories today
- Simile AI raises $2B Series B for human behavior simulation
- ChatGPT adds recent photos shortcut and time features
- Best GPU Neoclouds 2026: CoreWeave, Nebius, Lambda, Crusoe, Groq Ranked
- Anthropic's Opus 4.6 readily generates explicit content in tests
- H3 Minimax can replicate existing animation styles