DeepSeek V4.1 Flash 1017 is the October 17, 2026 snapshot of DeepSeek V4.1 Flash, served under the `deepseek-flash` API model name. It is a sparse mixture-of-experts model built on DeepSeek's Causal Encoder-Decoder (CED) architecture, activating 8B parameters on input and 16B on output from a 552B-parameter backbone. This asymmetric split keeps per-token compute low relative to total model size.
Image understanding is native, with visual and text embeddings trained jointly from pre-training start. The model suits coding, terminal, computer-use agents, long-horizon tasks, and long-context analysis. Compressed KV caching reduces cache memory to roughly a quarter of the prior Flash generation, lowering costs for agentic workloads.
DeepSeek positions it as the cost-efficient tier of the V4.1 family, reporting it exceeds V4 Pro in performance, speed, and task completion time. Pricing is $0.15 per 1M input tokens and $0.60 per 1M output tokens, with a 1.0M token context length. Hosted exclusively via Open Router, which forwards requests directly without routing decisions.
Performance metrics show 144 tokens per second throughput, 1.29s latency (P50), 99.79% availability over 3 days, and 100.00% uptime. The model is not currently available through EU in-region routing.
Source: Hacker News · Summarized by HeadlinesBriefing