Skip to content
OfferSign up and get $10 of credit for any model in Chat and the API. Top up from $1, with no processing fees. See free models →·Some free models are at the daily cap — refills 05:30 IST; requests auto-route to a live backup. Status →
One AI API Key

DeepSeek · DeepSeek · Balanced

DeepSeek V4 Pro (free lane)

New

DeepSeek's larger V4 Pro on the free lane for harder coding and reasoning, text input only.

Model id

deepseek-v4-pro:free

Price per 1M tokens

Input
$0.49
Output
$0.98
Cached input
Not offered

Added here 11 Nov 2026 — read the changelog for what changed and why.

Output is 2× the input rate. $1 covers about 1,020 prompts of 1,000 input + 500 output tokens. 1 credit = $1 of usage. All rates

Start using DeepSeek V4 Pro (free lane)

Set ONEAIAPIKEY_API_KEY to a key from your console, then run any example. Streaming is on so long answers are not cut off by idle-connection timeouts.

curl https://api.oneaiapikey.com/v1/chat/completions \
  -H "Authorization: Bearer $ONEAIAPIKEY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-pro:free",
    "stream": true,
    "messages": [{"role": "user", "content": "Make the failing test in checkout.test.ts pass without changing the test."}]
  }'
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.oneaiapikey.com/v1",
    api_key=os.environ["ONEAIAPIKEY_API_KEY"],
)

stream = client.chat.completions.create(
    model="deepseek-v4-pro:free",
    messages=[{"role": "user", "content": "Make the failing test in checkout.test.ts pass without changing the test."}],
    stream=True,  # recommended for long outputs
)
for chunk in stream:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.oneaiapikey.com/v1",
  apiKey: process.env.ONEAIAPIKEY_API_KEY,
});

const stream = await client.chat.completions.create({
  model: "deepseek-v4-pro:free",
  messages: [{ role: "user", content: "Make the failing test in checkout.test.ts pass without changing the test." }],
  stream: true,
});

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? "");
}

More: quickstart · API reference · try it in Chat first

Tokens per day, by model

Last 7 days, UTC days

Hover, tap or drag across any day, or focus the chart and use the arrow keys, to see how every model did that day.

Real tokens served through this gateway, updated every 5 minutes. Ranked by usage, not quality.

Full ranking insights

DeepSeek line-up

DeepSeek V4 Pro (free lane) or DeepSeek V4 Pro or Mistral Large 4?

Of the 4 DeepSeek models on this site, DeepSeek V4 Pro (free lane) is the top of the line-up and DeepSeek V4 Pro the step down. Mistral Large 4 from Mistral AI is the closest-priced balanced-tier model from another vendor. Each comparison lists the published rates and limits, then what favours each side.

DeepSeek V4 Pro (free lane)vsDeepSeek V4 Pro

Per 1M tokensDeepSeek V4 Pro (free lane)DeepSeek V4 Pro
Input$0.49the same$0.49
Output$0.98the same$0.98
Context1Mthe same1M
Max output384kthe same384k

Pick DeepSeek V4 Pro (free lane) when

Nothing in the published data puts it ahead of DeepSeek V4 Pro.

Pick DeepSeek V4 Pro instead when

  • DeepSeek V4 Pro (free lane) does not list Prompt caching and your code depends on it.

DeepSeek V4 Pro (free lane)vsMistral Large 4

Per 1M tokensDeepSeek V4 Pro (free lane)Mistral Large 4
Input$0.493% more$0.476
Output$0.9831% less$1.43
Context1M91% larger524.3k
Max output384k46% more262.1k

Pick DeepSeek V4 Pro (free lane) when

  • cost per request decides it: DeepSeek V4 Pro (free lane) charges output 31% less than Mistral Large 4.
  • prompts run past 524.3k tokens: DeepSeek V4 Pro (free lane) reads up to 1M.
  • one answer must be longer than 262.1k tokens (its limit is 384k).
  • you rely on JSON mode, which Mistral Large 4 does not list.
  • the work is coding agents, hard reasoning and math, which this catalogue lists for DeepSeek V4 Pro (free lane) and not for Mistral Large 4.

Pick Mistral Large 4 instead when

  • the job runs well enough on the lower rates: input at 3% less.
  • DeepSeek V4 Pro (free lane) does not list Vision, Structured output and Prompt caching and your code depends on it.
  • you send images, which only Mistral Large 4 of the two accepts.
  • your workload looks like coding, tool use and structured output, labels the catalogue gives Mistral Large 4 only.

Use cases

What DeepSeek V4 Pro (free lane) is listed for, and what it costs

The catalogue lists DeepSeek V4 Pro (free lane) for coding agents, hard reasoning and math. Each row prices one typical request at the published rates, with the arithmetic shown. Token counts are examples; your prompts will differ.

Model your own traffic

  1. Coding agents

    One step of an agent loop that edits files and runs tests: 40,000 tokens of context, and 2,500 tokens out.

    Works well as the main model in an agent loop that edits files and runs tests.

    “Make the failing test in checkout.test.ts pass without changing the test.”

    Same request on DeepSeek V4 Pro $0.006138 and Mistral Large 4 $0.011186. Cheapest balanced-tier model in the family: DeepSeek V4 Pro at $0.006138, 72% less.

    $0.02205

    per agent step, about 453 per $10

    0.04 × $0.49 + 0.0025 × $0.98

  2. Hard reasoning

    A multi-step problem: 5,000 tokens in, 12,000 tokens out, most of it reasoning.

    Reserve it for problems where a cheaper model has already struggled.

    “Find the bug in this lock-free queue and prove the fix is correct.”

    Same request on DeepSeek V4 Pro $0.01421 and Mistral Large 4 $0.019516. No other balanced-tier DeepSeek model on this site does this job for less.

    $0.01421

    per problem, about 703 per $10

    0.005 × $0.49 + 0.012 × $0.98

  3. Math

    A calculation or proof with every step shown: 1,000 tokens in, 6,000 tokens out.

    Ask it to show intermediate steps so results can be checked.

    “Show step by step whether the series of 1/(n log n) converges.”

    Same request on DeepSeek V4 Pro $0.00637 and Mistral Large 4 $0.009044. No other balanced-tier DeepSeek model on this site does this job for less.

    $0.00637

    per problem, about 1,569 per $10

    0.001 × $0.49 + 0.006 × $0.98

Overview

When DeepSeek V4 Pro (free lane) fits

DeepSeek V4 Pro (free lane) is a balanced model from DeepSeek, available here under the exact model id “deepseek-v4-pro:free”. It heads the DeepSeek line-up here, above DeepSeek V4 Pro, DeepSeek V4 Flash (free lane) and DeepSeek V4 Flash. DeepSeek has not published a release date we can cite; it has been available here since 11 Nov 2026.

DeepSeek documents tools, reasoning and json mode for it.

It costs $0.49 per million input tokens and $0.98 per million output tokens, so output is 2× the input rate. On the use cases below, output makes up 11% of the bill for coding agents, 83% for hard reasoning and 92% for math, so capping max tokens and asking for diffs saves more than trimming the prompt. With a 3:1 input-to-output mix it ranks 10th lowest-cost of the 27 models on this site.

There is no discounted cached-input rate for this model, so every input token bills at $0.49 per million.

It is served through the OpenAI-compatible format at https://api.oneaiapikey.com/v1, so any SDK or tool that lets you set a custom base URL, for example Codex CLI, Cursor and Cline, can use it. It is not offered on the Anthropic Messages endpoint; pick an Anthropic-format model for Claude Code.

Use DeepSeek V4 Pro (free lane) with any balance.

DeepSeek V4 Pro (free lane) is a paid model: any balance above $0 pays for it, the $10.00 sign-up credit included. When it runs out, add $1 or more to your wallet. Every request is billed at the rates above, with a receipt in your usage log.

DeepSeek V4 Pro (free lane) is developed by DeepSeek. Model names and trademarks belong to their owners. One AI API Key is an independent API gateway and is not affiliated with, endorsed by or sponsored by DeepSeek.