Skip to main content
deepseek-r1 is DeepSeek’s R1-series reasoning model — explicit chain-of-thought output formatted so downstream consumers can show, hide, or summarize the reasoning track separately from the final answer. The right pick for math, multi-step deduction, code analysis, and anywhere a model’s reasoning trace needs to be auditable rather than implicit. Pricing: see the rate card. Failures don’t bill.

​
Protocols

ProtocolPath
OpenAI Chat CompletionsPOST https://llm.bytespike.ai/v1/chat/completions

​
Quickstart

curl https://llm.bytespike.ai/v1/chat/completions \
  -H "Authorization: Bearer $BYTESPIKE_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "model": "deepseek-r1",
    "messages": [
      { "role": "user", "content": "If a train leaves Beijing at 8:00 traveling 200 km/h, and another leaves Shanghai at 9:00 at 250 km/h on the same track, when do they meet? Beijing-Shanghai = 1200 km." }
    ],
    "reasoning_effort": "high"
  }'
The response includes the reasoning chain alongside the final answer; clients can render them separately or fold the chain into a collapsible UI.

​
Capabilities

CapabilitySupported
Chat Completions✅
Streaming (SSE)✅
Explicit reasoning chain✅
reasoning_effort low / medium / high✅
Tools / function calling✅
JSON mode✅
Vision (image input)—
Context window64K tokens
Modalitychat
Capability bucketexternal_chat

​
When to use

  • Math / multi-step deduction — where the chain matters for verification.
  • Code analysis — bug-trace style problems where intermediate reasoning helps the user (or another agent) audit the conclusion.
  • DOSIA chat_with — main brain calls this when the user asks “reason through” or “step by step” prompts.
When not to use:
  • Simple chat or summarization — deepseek-v3-2 is faster and cheaper without the reasoning premium.
  • Vision tasks — use gemini-3-1-pro.

​
Next