DeepSeek
Models published by DeepSeek, available through the AnyRouter API. Each can route across multiple upstream providers for availability and price.
DeepSeek's flagship reasoning model with chain-of-thought. Strong math, coding, and complex problem solving.
DeepSeek V3.1 is a hybrid reasoning model from DeepSeek with a 131,072-token context window, strong coding and agentic tool-use performance. Served free through SambaNova Cloud's fast RDU inference and SiliconFlow.
671B MoE general-purpose model with strong multilingual, coding, and reasoning performance.
Fast and cheap DeepSeek-V4 MoE variant via SiliconFlow free tier for high-throughput chat and coding.
DeepSeek V4 Pro is a high-capability reasoning model from DeepSeek with a 1M-token context window on Workers AI, plus BYOK routes for other hosts.
DeepSeek V4.1 Flash 1017 is the October 17, 2026 snapshot of DeepSeek V4.1 Flash, served by DeepSeek under the `deepseek-flash` API model name. Sparse mixture-of-experts with native vision in, text out, and a 1M-token context window. Hosts bill it as its own model: OpenRouter lists it at $0.15 / $0.60 per 1M input / output tokens, against $0.04 / $0.144 for the original September checkpoint `deepseek/deepseek-v4.1-flash`.
DeepSeek V4.1 Flash is the cost-efficient sparse MoE tier of the V4.1 family, built on the Causal Encoder-Decoder (CED) architecture: 552B total parameters with 8B active on input and 16B on output. Native vision in, text out, 1M context. Compressed KV cache needs roughly a quarter of the previous Flash's HBM. This is the original September 2026 checkpoint, listed by OpenRouter as `deepseek/deepseek-v4.1-flash`. The official API id `deepseek-flash` now serves the October 17 snapshot, which has its own listing `deepseek/deepseek-v4.1-flash-1017`.