DeepSeek V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the first built on the company's Causal Encoder-Decoder (CED) architecture. It activates 8B parameters on input and 16B on output from a 552B-parameter backbone, an asymmetric split that keeps per-token compute low relative to the model's total size. Image understanding is native to the architecture, with visual and text embeddings trained jointly from the start of pre-training rather than added afterward as in the earlier experimental V4 Flash Vision Exp(opens in new tab).
- By:
- DeepSeek
- Input:
- Output:
- Context length:
- 1M
- Max output:
- 384K
- Published:
- 2026-09-10
Providers
Same model, different providers. Automatic routing spreads requests across healthy providers by default, or ranks them by price, latency and more.
Availability
24 hoursSuccess rate of requests sent by the Amux AI gateway to each provider. Client errors are excluded.
Tok/s
24 hoursAverage output speed of streaming requests, in tokens per second (TPS).
TTFT
24 hoursAverage time to first token (TTFT) of streaming requests, measured from request dispatch to the first token received.
Activity
24 hoursRelative request activity over time, broken down by serving provider. Indexed to the peak within the selected range (= 100); reflects trends and provider share, not absolute volume.
Related models
More models from DeepSeek