DeepSeek-V4.1-Flash

POST /v1/chat/completions model: DeepSeek-V4.1-Flash

Specification

Architecture	DeepseekV41ForCausalLM, model_type = deepseek_v41
Layers	40
Hidden dimension	5,120
Attention	MLA — 64 query heads, 1 KV head, head dim 512, RoPE dim 64
Mixture of Experts	384 routed + 1 shared, 6 routed activated, expert intermediate 2,304
Sparse-attention indexer	32 heads, head dim 128, top-512 tokens
Engram	16,000,000 n-grams (compressed to 99,092), 8 heads, max n-gram 4
Multi-token prediction	3 layers
Vocabulary	129,280
Vision tower	32 layers, width 1,024, 16 heads, patch 14, max_image_tokens = 1024
Quantisation	FP8 weights, block [32, 32], ue8m0 scales; experts in FP4

At a glance

Context length	1,048,576 tokens
Input modalities	text + images
Streaming	✅
Tool calling	✅
JSON output	✅ json_object
Thinking	✅ — off by default
Inference engine	SGLang
Stability	experimental

Thinking

Omit reasoning_effort and the model answers directly, with no thinking pass.

Every tier is accepted: none, minimal, low, medium, high, xhigh, max.

Thinking text arrives in choices[0].message.reasoning, and the token count in usage.completion_tokens_details.reasoning_tokens.

messages

system messages may appear at any position, and there may be more than one. Both system and developer are accepted as the role name.

Image input

Send an image_url content part:

{
  "model": "DeepSeek-V4.1-Flash",
  "messages": [{
    "role": "user",
    "content": [
      { "type": "image_url", "image_url": { "url": "data:image/png;base64,iVBORw0KGgo..." } },
      { "type": "text", "text": "What does this image say?" }
    ]
  }]
}

Image usage is reported at usage.prompt_tokens_details.image_tokens. The weights cap a single image at 1,024 tokens.

Limits

Context window	1,048,576 tokens, prompt plus output
JSON output	response_format: {"type": "json_object"}
Thinking tiers	none minimal low medium high xhigh max

Exceeding the window returns:

Requested token count exceeds the model's maximum context length of 1048576 tokens.

Example

curl https://developer.amd.com.cn/radeon/api/v1/chat/completions \
  -H "Authorization: Bearer $RADEON_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "DeepSeek-V4.1-Flash",
    "reasoning_effort": "medium",
    "messages": [
      { "role": "user", "content": "Why is the sky blue?" }
    ]
  }'