WHAT NEW IN X10.28 WED

AI collections from X · 22 collections · 58,924 posts

Back to No.020 DeepSeek V4.1 Flash hands-on tests and comparisonsUpdated 7 hours ago
New todaydeepseek-v4.1IMAGE

TinyRouter pitches DeepSeek V4.1 Flash 1017's 1M context over retrieval pipelines

TinyRouter says DeepSeek-V4.1-Flash-1017 is live with a 1M-token context for lightweight reasoning and high throughput, arguing that if a file fits you should send it whole instead of chunking and reranking.

TinyRouter
@TinyRouterHQArchived 8 hours agoOriginally posted 2026.10.28
TinyRouter pitches DeepSeek V4.1 Flash 1017's 1M context over retrieval pipelines (image 1 of 4)
TinyRouter pitches DeepSeek V4.1 Flash 1017's 1M context over retrieval pipelines (image 2 of 4)
TinyRouter pitches DeepSeek V4.1 Flash 1017's 1M context over retrieval pipelines (image 3 of 4)
TinyRouter pitches DeepSeek V4.1 Flash 1017's 1M context over retrieval pipelines (image 4 of 4)
TinyRouter @TinyRouterHQOriginally posted 2026.10.28
Open original on X

Retrieval pipelines exist because the context window was small. DeepSeek-V4.1-Flash-1017 is live: 1M-token context for lightweight reasoning and high-throughput work. If the whole file fits, send it whole: overlap and rerank add failure modes. #AgenticAI

x.com/tinyrouterhq/status/2115281799517794778Archived 2026.10.28 12:02
Open original on X