About DeepSeek V4.1 Flash
DeepSeek V4.1 Flash is a natively multimodal Mixture-of-Experts model built for coding, reasoning, and long-horizon agent tasks. It is the cost-efficient tier of the V4.1 family, and DeepSeek reports it ahead of V4 Pro on quality, speed, cost, and total task completion time, with V4 Flash and V4 Pro both retired in its favor.
The model uses DeepSeek's Causal Encoder-Decoder architecture, which activates 8B parameters while reading input and 16B while generating output, and compresses the KV cache to roughly a quarter of V4 Flash's footprint. It reads images and text through a DeepSeek-ViT vision encoder and supports a 1M-token context window, tool use, structured outputs, and a continuously adjustable reasoning-effort setting (1-100) that trades inference cost for accuracy. Pretraining covers 45T multimodal tokens, followed by SFT, RL, and on-policy distillation post-training. Weights are published under the MIT license.
| Metric | Value |
|---|---|
| Parameter Count | 552 billion |
| Mixture of Experts | Yes |
| Active Parameter Count | 8 billion (prefill) / 16 billion (decode) |
| Context Length | 1,048,576 tokens |
| Max Output | 384,000 tokens |
| Multilingual | Yes |
| Tool Use | Yes |
| Structured Outputs | Yes |
Ready to build with DeepSeek V4.1 Flash?
Try DeepSeek V4.1 Flash in the Workbench to prompt it, compare outputs, and iterate on prompts without writing any code. When you're ready to ship, call the same model from our API and build your own apps on top of it.
curl -sSf -X POST https://hub.oxen.ai/api/ai/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $OXEN_API_KEY" \
-d '{
"model": "deepseek-v4-1-flash",
"messages": [
{
"role": "user",
"content": "Try sending a message."
}
]
}'API endpoint
See the API reference for request and response formats.
https://hub.oxen.ai/api/ai/chat/completions