DeepSeek V4 — free API (Pro & Flash), OpenAI-compatible

DeepSeek V4 is one of the strongest open-weight model families for reasoning, coding, and agentic work — and through InferAll you can call it **free** via NVIDIA NIM. Both tiers are available: `deepseek-ai/deepseek-v4-pro` for maximum capability and `deepseek-ai/deepseek-v4-flash` for cost-efficient, lower-latency work. No credit card, $0 within the free tier, and it works with the OpenAI SDK you already have. | Model | Best for | |---|---| | `deepseek-ai/deepseek-v4-pro` | Maximum reasoning, coding, agentic tasks | | `deepseek-ai/deepseek-v4-flash` | Cost-efficient, lower-latency, high-volume | Both are $0 on the free NVIDIA NIM tier. --- ### Quick start (Python) ```python from openai import OpenAI client = OpenAI( base_url="https://api.inferall.ai/v1", api_key="ifu_your_key_here", # get one at inferall.ai/keys — no card required ) # Flash — fast, cost-efficient, great default response = client.chat.completions.create( model="deepseek-ai/deepseek-v4-flash", messages=[{"role": "user", "content": "Refactor this function for readability: ..."}], max_tokens=512, ) # Pro — for the hardest reasoning / agentic tasks response = client.chat.completions.create( model="deepseek-ai/deepseek-v4-pro", messages=[{"role": "user", "content": "Design a rate limiter with a sliding window. Explain the tradeoffs."}], max_tokens=1024, ) print(response.choices[0].message.content) ``` The only change from calling OpenAI directly is the `base_url`. Your existing code, streaming, and tool use all work unchanged. --- ### TypeScript ```typescript import OpenAI from "openai"; const client = new OpenAI({ baseURL: "https://api.inferall.ai/v1", apiKey: process.env.INFERALL_API_KEY, }); const response = await client.chat.completions.create({ model: "deepseek-ai/deepseek-v4-flash", messages: [{ role: "user", content: "Write a SQL migration to add a nullable column." }], max_tokens: 400, }); console.log(response.choices[0].message.content); ``` --- ### Why call DeepSeek V4 through InferAll **It's genuinely $0 in / $0 out.** DeepSeek V4 runs on NVIDIA NIM (NVIDIA Inference Microservices) on their DGX Cloud infrastructure, which InferAll exposes at $0. No inference cost to pass through, so it stays $0 per call within the free-plan daily request caps (100 chat / 50 text / day, reset 00:00 UTC) — activate via the $5 starter pack at [/billing](https://inferall.ai/billing). **One key for DeepSeek and everything else.** The same `ifu_...` key also calls Qwen 3.5, GLM-5.1, Kimi K2.6, Llama 3.1/3.3 70B, and 110+ more free open models — plus paid GPT-4.1, Claude Opus 4, and Gemini 2.5 when you need a frontier model. Switch by changing one string; no juggling provider keys. **OpenAI-compatible.** Standard `chat.completion` responses, streaming, tool use, and JSON mode — all working with whatever OpenAI client you already have. Moving from `gpt-4o-mini` to `deepseek-ai/deepseek-v4-flash` is a one-line model-string change. --- ### Pro vs Flash — which to use Start with **Flash**. It handles the majority of coding, refactoring, summarization, and structured-output tasks at lower latency, and it's free. Step up to **Pro** when you hit a genuinely hard reasoning or multi-step agentic problem where you can feel Flash struggling. Since both are $0 on the free tier, the only real cost of using Pro is latency — so use Flash by default and reserve Pro for the hard cases. --- ### Compare it yourself The best way to pick a model is to watch several answer the same prompt. The full free roster is one call away — `curl https://api.inferall.ai/ai/v1/models` — so you never hardcode a list that goes stale. Sign up at [inferall.ai/keys](https://inferall.ai/keys), then activate via the $5 starter pack at [/billing](https://inferall.ai/billing) — the $5 becomes spendable balance for premium providers and unlocks the 118+ NIM open models at $0 in/out (within the free-plan daily request caps).