DeepSeek unveils V4.1-Flash: new encoder-decoder architecture with native vision
gaganghotra_ · x · 2026-09-11
DeepSeek officially introduced DeepSeek-V4.1-Flash, the smallest model in its new architecture family, with native visual understanding built in. The release marks a major overhaul of the V4.1 line around an encoder-decoder setup, aimed at higher capability, faster inference, and greater throughput — designed as a foundation that scales to larger models.
Researcher Sebastian Raschka (rasbt) called it a big overhaul so substantial that "they should have called it DeepSeek V5," praising the return of the encoder-decoder design as refreshing. Detailed benchmarks and technical specifics are expected in the follow-up posts of the official thread.
More from Models
- Users notice o5 constantly self-measures while f5/5.1 barely does — repligate · 2026-09-11
- Chinese labs' distillation scale revealed: Alibaba 151M+ exchanges, Moonshot 23M+, DeepSeek 12.1M+ — scaling01 · 2026-09-11
- Astra usage limits worse than Fable: burn a week's quota in a single day — cocktailpeanut · 2026-09-11
- DeepSeek V4.1 Flash Architecture: 552B MoE with Asymmetric 8B Read / 16B Decode Compute — demian_ai · 2026-09-11
- Insider: chances are zero that opt-in Codex data made it into OpenAI training runs — soumitrashukla9 · 2026-09-11
- FrontierMath Tier 4 fully solved: GPT-6 Astra cracks the last problem standing — Jsevillamol · 2026-09-11