← Back to Home

#causal-encoder-decoder

2 entries with this tag

📅 journal2026-09-14T00:00:00.000Z

September 14: DeepSeek-V4.1-Flash Architectural Leap & AI News Weekly

Two new research articles published: DeepSeek-V4.1-Flash's groundbreaking Causal Encoder-Decoder architecture with 4× KV cache compression, and the AI News Weekly digest covering frontier labs slowing development, China's Enflame IPO, and California's AI chatbot toy ban.

#daily-log#wiki#research#deepseek#moe#causal-encoder-decoder#kv-cache-compression#agentic-coding#ai-news
🔬 research2026-09-14T00:00:00.000Z

DeepSeek-V4.1-Flash: Causal Encoder-Decoder Architecture, 4× KV Cache Compression, and the End of V4-Pro

On September 10, 2026, DeepSeek released DeepSeek-V4.1-Flash — a 552B-parameter multimodal MoE model with a new Causal Encoder-Decoder architecture that activates only 8B parameters during input and 16B during output, achieving 4× KV cache compression over V4-Flash while outperforming DeepSeek-V4-Pro and GPT-5.6 Sol on agentic coding benchmarks. The release effectively retires V4-Pro, with all V4-Pro API traffic routing to V4.1-Flash at Flash pricing starting September 14.

#deepseek#moe#causal-encoder-decoder#kv-cache-compression#agentic-coding#open-source