Deepseek OCR 2
oah/deepseek-v3.1Deploy Deepseek OCR 2 with built-in PII redaction and Hub governance. Available on Managed Credits and BYOK.
Together.aiDeepInfra
Input: $0.21/MOutput: $0.79/M
Why deploy DeepSeek through AI Security Gateway?
Automatic PII Redaction
Every DeepSeek request is scanned for 28+ PII entity types — SSNs, credit cards, emails, API keys, and more — before it reaches any provider.
Smart Cost Routing
DeepSeek is available across 2 providers. Our Smart Router picks the cheapest one per-request. 25% managed markup / 0% on Pro BYOK.
Native SDK or OpenAI Compatible
Use the AISG SDK (pip install aisg) for typed metadata and error handling, or change two lines in your OpenAI SDK. Both work.
Full Observability
Per-request logging of token counts, latency, DLP violations, and cost. Never wonder what your AI spend is again.
DeepSeek Strengths
- Top-tier coding performance rivaling GPT-4 on benchmarks
- DeepSeek R1: strong chain-of-thought reasoning
- Extremely cost-effective for its capability level
- Open-weights for auditability and fine-tuning
- Growing provider support across Groq, Together, DeepInfra
Available DeepSeek Models (11)
DeepSeek R1 0528 NVFP4
oah/deepseek-r1Deploy DeepSeek R1 0528 NVFP4 with built-in PII redaction and Hub governance. Available on Managed Credits and BYOK.
Together.ai
ReasoningInput: $3.00/MOutput: $7.00/M
DeepSeek R1 Distill Qwen 1.5B
oah/deepseek-r1-distill-qwen-1.5bDeploy DeepSeek R1 Distill Qwen 1.5B with built-in PII redaction and Hub governance. Available on Managed Credits and BYOK.
Together.ai
ReasoningInput: $0.18/MOutput: $0.18/M
DeepSeek R1 Distill Qwen 14B
oah/deepseek-r1-distill-qwenDeploy DeepSeek R1 Distill Qwen 14B with built-in PII redaction and Hub governance. Available on Managed Credits and BYOK.
Together.ai
ReasoningInput: $1.60/MOutput: $1.60/M
Deepseek V4 Pro
oah/deepseek-v4-proDeploy Deepseek V4 Pro with built-in PII redaction and Hub governance. Available on Managed Credits and BYOK.
Together.aiDeepInfra
Input: $2.10/MOutput: $4.40/M
Deepseek Coder 33B Instruct
oah/deepseek-coderDeploy Deepseek Coder 33B Instruct with built-in PII redaction and Hub governance. Available on Managed Credits and BYOK.
Together.ai
Input: $0.80/MOutput: $0.80/M
deepseek-ai/DeepSeek-R1-0528
oah/deepseek-r1-0528Deploy deepseek-ai/DeepSeek-R1-0528 with built-in PII redaction and Hub governance. Available on Managed Credits and BYOK.
DeepInfra
ReasoningInput: $0.50/MOutput: $2.15/M
deepseek-ai/DeepSeek-V3
oah/deepseekDeploy deepseek-ai/DeepSeek-V3 with built-in PII redaction and Hub governance. Available on Managed Credits and BYOK.
DeepInfra
Input: $0.25/MOutput: $0.88/M
deepseek-ai/DeepSeek-V3.1-Terminus
oah/deepseek-v3.1-terminusDeploy deepseek-ai/DeepSeek-V3.1-Terminus with built-in PII redaction and Hub governance. Available on Managed Credits and BYOK.
DeepInfra
Input: $0.27/MOutput: $1.00/M
deepseek-ai/DeepSeek-V3.2
oah/deepseek-v3.2Deploy deepseek-ai/DeepSeek-V3.2 with built-in PII redaction and Hub governance. Available on Managed Credits and BYOK.
DeepInfra
Input: Free/MOutput: Free/M
deepseek-ai/DeepSeek-V4-Flash
oah/deepseek-v4-flashDeploy deepseek-ai/DeepSeek-V4-Flash with built-in PII redaction and Hub governance. Available on Managed Credits and BYOK.
DeepInfra
Input: Free/MOutput: Free/M
DeepSeek Pricing Comparison (per 1M tokens, USD)
Input / Output pricing by provider. Managed Mode adds a 25% managed markup. Pro BYOK = 0% markup.
| Model | Params | Context | Vision | Together.ai | DeepInfra |
|---|---|---|---|---|---|
Deepseek OCR 2 oah/deepseek-v3.1 | — | 8K | No | $0.60/$1.70 | $0.21/$0.79 |
DeepSeek R1 0528 NVFP4 oah/deepseek-r1 | — | 164K | No | $3.00/$7.00 | — |
DeepSeek R1 Distill Qwen 1.5B oah/deepseek-r1-distill-qwen-1.5b | — | 131K | No | $0.18/$0.18 | — |
DeepSeek R1 Distill Qwen 14B oah/deepseek-r1-distill-qwen | — | 131K | No | $1.60/$1.60 | — |
Deepseek V4 Pro oah/deepseek-v4-pro | — | 512K | No | $2.10/$4.40 | — |
Deepseek Coder 33B Instruct oah/deepseek-coder | — | 16K | No | $0.80/$0.80 | — |
deepseek-ai/DeepSeek-R1-0528 oah/deepseek-r1-0528 | — | — | No | — | $0.50/$2.15 |
deepseek-ai/DeepSeek-V3 oah/deepseek | — | — | No | — | $0.25/$0.88 |
deepseek-ai/DeepSeek-V3.1-Terminus oah/deepseek-v3.1-terminus | — | — | No | — | $0.27/$1.00 |
deepseek-ai/DeepSeek-V3.2 oah/deepseek-v3.2 | — | — | No | — | — |
deepseek-ai/DeepSeek-V4-Flash oah/deepseek-v4-flash | — | — | No | — | — |
DeepSeek Direct vs AI Security Gateway
What you get at each pricing tier. Hub adds security, governance, and multi-provider routing on top of raw API access.
| Mode | What You Pay | PII Redaction | Budget Caps | Routing | Audit Trail |
|---|---|---|---|---|---|
| Direct to DeepSeek | Provider pricing only | None | None | Manual | None |
| Hub — Managed Mode | Provider + 25% markup | 28+ PII types | Per-key hard caps | Smart Router | Full compliance log |
| Hub — Pro BYOK ($29/mo) | Direct to provider (0% markup) | 28+ PII types | Per-key hard caps | Smart Router | Full compliance log |
Popular Use Cases
1
Code generation, review, and debugging
2
Technical documentation and analysis
3
Math and logical reasoning tasks
4
Cost-optimized production workloads replacing GPT-4 class models
Quick Integration
# pip install aisg
from aisg import AISG
client = AISG(api_key="your_hub_api_key")
response = client.chat.create(
model="oah/deepseek-v3.1",
messages=[{"role": "user", "content": "Hello!"}],
)
print(response.content)
print(response.aisg_metadata.pii_detected)
print(response.aisg_metadata.cost_usd)Use any virtual model name from the pricing table above (prefixed with oah/). Also works with the standard OpenAI SDK — just change base_url. Every request is PII-scanned before reaching DeepSeek (Open Source).
Frequently Asked Questions
What is the DeepSeek API pricing?
DeepSeek API pricing varies by provider and model variant. In Managed Mode, we add a 25% markup. With Pro BYOK, you pay the provider directly at 0% markup. Check the pricing table above for live rates across Groq, Together.ai, and DeepInfra.
What is the DeepSeek V3 cost?
DeepSeek V3 cost depends on the provider. It's one of the most cost-effective models at its capability level — see the pricing comparison table above for per-token rates.
How do I access the DeepSeek API?
Through AI Security Gateway, use the standard OpenAI SDK with model name 'oah/deepseek-v3' or 'oah/deepseek-r1'. All requests get 28+ PII entity redaction automatically.
How does DeepSeek compare to GPT-4?
DeepSeek V3 matches or exceeds GPT-4 on many coding and reasoning benchmarks at a significantly lower cost. For coding tasks specifically, it's one of the best options available.
Is DeepSeek available in Managed Mode?
Yes. DeepSeek models are available through multiple providers in both Managed and BYOK modes.
Previous Versions
These models have been retired by the provider. Migrate to a current variant above.
DeepSeek R1-0528together
DeepSeek R1 0528together
DeepSeek R1 (Original)together
deepseek-ai/DeepSeek-V3together
Deepseek V3 0324together
Deepseek V3 Basetogether
Deepseek V3together
Deepseek V3.1together
Deepseek V3.1 Basetogether
Deepseek V3.1 Terminustogether
Deepseek V3.2together
Deepseek V3.2 Exptogether
deepseek-ai/DeepSeek-OCRdeepinfra
deepseek-ai/DeepSeek-R1-0528-Turbodeepinfra
deepseek-ai/Janus-Pro-1Bdeepinfra
deepseek-ai/Janus-Pro-7Bdeepinfra
Deploy DeepSeek with Enterprise-Grade Security
Get started with 1,000,000 free credits. Every DeepSeek request is PII-scanned, cost-optimized, and fully logged — zero configuration.
Not ready yet? Get notified about DeepSeek updates:
Explore Other Model Families
🦙Llama
Meta's open-weights Llama family is the most widely deployed open-source LLM series. Compare Llama API pricing across Gr…
🧠GPT
OpenAI's GPT family powers the majority of commercial AI applications. Compare GPT-4 API cost and OpenAI API pricing acr…
💎Gemini
Google's Gemini family offers powerful multimodal capabilities with large context windows. Compare Gemini API pricing an…
🤖Claude
Anthropic's Claude family is built with safety and reliability at its core. Compare Claude API pricing and Claude Sonnet…
🌀Mistral
Mistral AI's model family spans from compact open-weights models to powerful commercial variants. Compare Mistral API pr…
Model registry last updated: . Pricing shown is the lowest available rate across providers (per 1M tokens, USD). Actual pricing depends on provider and plan.