the closed line is coming down to meet us. We're not going up to meet them. Thank you, Deep Seekers V4 flash and kimi k 3
Summary
The author observes that the gap between closed and open AI models is narrowing, crediting DeepSeek V4 Flash and Kimi K3 for bringing the closed line down to meet the open-source community.
Similar Articles
@QuixiAI: Amazing, @Kimi_Moonshot! I guess we don't need @deepseek_ai v4 anymore. (AI simply moves too fast to sit on your hands.)
Social media post suggests Kimi Moonshot's latest model leapfrogs expectations for DeepSeek v4.
Open vs Closed AI Models: How the Gap Collapsed in 2025-2026 and Where It's Heading
The article examines how the performance gap between open and closed AI models has narrowed dramatically from early 2025 to mid-2026, highlighted by DeepSeek's open model release and the subsequent market impact. It discusses the roles of Chinese labs in driving the open frontier and the implications for the industry.
Deepseek, kimi etc..
Mentions of AI models Deepseek and Kimi, possibly discussing recent updates or comparisons.
Deepseek V4 Flash is now ~#2 open weight model to Kimi K3 and >50x cheaper
DeepSeek's new V4 Flash model is reportedly the #2 open-weight model behind Kimi K3, offering strong performance at over 50x lower cost ($0.09/$0.18 per 1M tokens) with solid coding and reasoning capabilities.
We tested Deepseek v4 flash, GLM 5.2, and Kimi K3 on hard agentic tasks, and DeepSeek just crushed
Composio tested DeepSeek v4 flash, GLM 5.2, and Kimi K3 on hard agentic tasks, finding DeepSeek the fastest and cheapest with roughly the same success rate as the others, while frontier models still lead slightly.