DeepSeek-V3

Cost & access

DeepSeek-V3 weights are available under DeepSeek Model License. Direct API cost may be $0 if you self-host; budget for GPUs, storage, and engineering instead. Hosted endpoints (Together, Fireworks, Groq, etc.) charge per token—shop providers for deepseek-v3 latency and region. With a 128K tokens context window, long PDFs or chat histories increase input tokens quickly—trim history or summarize older turns in production.

Functional understanding

• Large MoE model competitive on reasoning and coding with open weights.
• Modalities: text · License: DeepSeek Model License · Released 2024-12-26.
• Best-fit workflows for this model:
• • Drafting, summarization, and structured extraction from long documents.
• • On-prem or VPC deployment when data cannot leave your network.

Technical foundation

• DeepSeek reports 671B MoE (~37B active per token, public report) parameters; training data: 14.8T tokens (technical report).
• Context: 128K tokens. Open weights: yes.
• DeepSeek-V3 is positioned as a general-purpose model in the DeepSeek lineup.

First API call

DeepSeek exposes an OpenAI-compatible API—set base_url and use model deepseek-chat.

from openai import OpenAI

client = OpenAI(api_key="YOUR_KEY", base_url="https://api.deepseek.com")
resp = client.chat.completions.create(
    model="deepseek-chat",
    messages=[{"role": "user", "content": "Hello from DeepSeek-V3"}],
)
print(resp.choices[0].message.content)

Important technical topics

• Prompting DeepSeek-V3: be explicit about output format. Weak: "Analyze this." Better: "Return JSON with fields id, total, date for DeepSeek billing data."
• Temperature: use 0–0.3 for extraction and compliance on DeepSeek-V3; 0.7–1.0 for brainstorming.
• Tokens: DeepSeek-V3 bills by tokens (~¾ word each). 671B MoE (~37B active per token, public report) parameters affect capability; your bill is driven by context length and call volume.
• Context window (128K tokens): everything in one request—system prompt, tools, RAG chunks, and history—must fit. Truncate or summarize when approaching the limit for DeepSeek-V3.

Real enterprise patterns

• RAG with DeepSeek-V3: retrieve from your vector DB, cite sources in the prompt.
• Tool calling: define JSON schemas; let DeepSeek-V3 request functions, not free-form SQL.
• Eval suite: regression prompts before each model or prompt change.
• Cost routing: default to DeepSeek-V3 for hard tasks; smaller sibling model for triage.

Production & security

• Secrets: never commit keys for DeepSeek-V3; use vault + per-environment rotation.
• PII: mask before inference; log redacted prompts only.
• Observability: trace id per request; log model=deepseek-v3, tokens in/out, latency.
• GPU monitoring: VRAM, batch queue depth, and model revision hash on each deploy.
• Guardrails: schema-validate JSON; block disallowed topics; cross-check numbers against source docs.

Mini projects with this model

• Support copilot: DeepSeek-V3 drafts replies from KB snippets.
• Contract clause extractor with human approval.
• Weekly metrics narrative from SQL + CSV exports.
• Agent that files expenses from receipt photos (if multimodal).

Suggested stack

• Language: Python 3.11+
• Model: DeepSeek-V3 via Ollama, vLLM, or Hugging Face
• Hardware: NVIDIA GPU with enough VRAM for quantization level
• API wrapper: FastAPI or LiteLLM proxy
• UI: Streamlit or Next.js for internal tools
• APIs: FastAPI
• Vector DB (RAG): Pinecone / Chroma / pgvector

Learning path

• Python basics
• HTTP/REST and environment variables
• DeepSeek authentication and DeepSeek-V3 model id (deepseek-chat)
• First successful call to DeepSeek-V3
• Prompt design and JSON / structured outputs
• RAG
• Tool use / function calling
• Evals and regression sets
• Production deploy + monitoring