API Need50% off
Uptime
Total Context
1.0486M
Max Output
384K
Input
$0.2175/ 1M
Output
$0.435/ 1M
Cache Read
$0.0018/ 1M
Cache Write
Route priced
DDeepSeek: DeepSeek V4 Pro API
deepseek/deepseek-v4-pro
DeepSeek V4 Pro API is a text-only reasoning model with roughly 1.05M context and the lowest output rate among ranked models in this catalog. It targets long, cost-sensitive analysis where latency is less critical. [1]
Its max run ranks #13 with a 49.97 median and 68.00 best score. The 1,468-second measured completion time is high, but the benchmark test cost is far below similarly ranked premium models. [1][2]
MODALITIES
Price$0.2175 / $0.435 / 1MCONTEXT1.0486M
Playground
Test DeepSeek V4 Pro API with representative production prompts before using deepseek-v4-pro in a live workflow.
Providers
APINEED lists DeepSeek V4 Pro API with text input, text output, and a 1.048576M context window. APINEED routing and prepaid rates are identified separately. [1]
Discount
APINEED applies a 50% discount to the current $0.435 input and $0.87 output official rates, producing $0.2175 and $0.435 per million tokens for batch jobs. The table compares official DeepSeek V4 Pro API pricing with the APINEED prepaid rate.
Official
Provider baselineInput $0.435/ 1MOutput $0.87/ 1M
BaselineAPINEED50% off
Prepaid creditsInput $0.2175/ 1MOutput $0.435/ 1M
Availability
APINEED continuously monitors DeepSeek V4 Pro API access and keeps requests on healthy capacity.
System statusLast 90 days
DeepSeek V4 Pro API Benchmarks
DeepSeek ranks just behind GPT-5.6 Luna at #13, with a slightly lower median and a much longer measured completion time. Its advantage is cost efficiency rather than speed. The complete LLM2014 logic 2026-07 table remains below, with DeepSeek V4 Pro API highlighted when a current row exists. [2]
| 1st | GPT-5.5 (xhigh) | 83.80 | 77.46 | 7.57% | 494s | 30,811 | $25.51 | $29.57 |
| 2nd | Kimi-K3 (max) | 82.91 | 74.80 | 9.78% | 1095s | 39,912 | $16.76 | $15.00 |
| 3rd | Claude Opus 4.8 (xhigh) | 82.62 | 66.70 | 19.27% | 791s | 41,833 | $28.86 | $24.64 |
| 4th | GPT-5.6 Sol (xhigh) | 81.94 | 72.99 | 10.92% | 369s | 17,433 | $14.43 | $29.57 |
| 5th | Claude Opus 5 (xhigh) | 78.38 | 71.88 | 8.29% | 426s | 26,392 | $18.21 | $24.64 |
| 6th | Qwen3.7-Max (xhigh) | 74.56 | 66.95 | 10.21% | 374s | 54,207 | $7.81 | $5.14 |
| 7th | GLM-5.2 (max) | 73.68 | 58.27 | 20.91% | 890s | 45,683 | $5.12 | $4.00 |
| 8th | Gemini 3.1 Pro (high) | 73.36 | 58.46 | 20.31% | 214s | 25,905 | $8.58 | $11.83 |
| 9th | GPT-5.6 Luna (xhigh) | 69.49 | 51.97 | 25.21% | 242s | 38,083 | $6.31 | $5.91 |
| 10th | DeepSeek V4 Pro (max) | 68.00 | 49.97 | 26.51% | 1468s | 60,558 | $1.45 | $0.86 |
| 11th | Grok 4.5 (high) | 67.59 | 56.72 | 16.08% | 451s | 43,372 | $7.18 | $5.91 |
| 12th | Muse Spark 1.1 | 66.18 | 57.23 | 13.52% | 542s | 42,416 | $4.98 | $4.19 |
| 13th | Gemini 3.5 Flash (high) | 65.92 | 60.39 | 8.39% | 188s | 36,372 | $9.03 | $8.87 |
| 14th | Doubao-Seed-2.1-pro (high) | 58.69 | 49.35 | 15.91% | 2003s | 85,059 | $10.21 | $4.29 |
| 15th | Qwen3.7-Plus (high) | 58.50 | 44.09 | 24.63% | 1237s | 57,152 | $1.83 | $1.14 |
| 16th | Gemini 3.6 Flash (high) | 56.87 | 41.74 | 26.6% | 206s | 26,329 | $5.45 | $7.39 |
| 17th | Tencent Hy3 (high) | 54.42 | 44.60 | 18.04% | 1035s | 52,037 | $0.83 | $0.57 |
| 18th | Claude Sonnet 5 (xhigh) | 51.37 | 43.14 | 16.02% | 529s | 44,118 | $12.18 | $9.86 |
| 19th | DeepSeek V4 Flash (max) | 50.62 | 36.24 | 28.41% | 611s | 49,593 | $0.40 | $0.29 |
| 20th | MiniMax-M3 | 49.18 | 40.94 | 16.75% | 985s | 56,467 | $1.90 | $1.20 |
| 21st | Claude Opus 5 | 42.08 | 31.73 | 24.6% | 154s | 10,180 | $7.02 | $24.64 |
| 22nd | Claude Opus 4.6 | 36.88 | 25.79 | 30.07% | 65s | 4,387 | $3.03 | $24.64 |
| 23rd | Doubao-Seed-2.0-lite 0428 (high) | 35.32 | 26.77 | 24.21% | 656s | 26,726 | $0.38 | $0.51 |
| 24th | Gemini 3.5 Flash (minimal) | 33.24 | 22.22 | 33.15% | 44s | 6,189 | $1.54 | $8.87 |
| 25th | Ling-3.0-flash | 32.53 | 20.62 | 36.61% | 688s | 86,204 | $0.00 | $0.00 |
| 26th | Gemini 3.5 Flash Lite (high) | 30.23 | 22.82 | 24.51% | 188s | 18,297 | $1.26 | $2.46 |
| 27th | GPT-5.5 Instant | 28.87 | 17.66 | 38.83% | 25s | 1,673 | $1.39 | $29.57 |
| 28th | Gemma 4 31B | 27.91 | 22.04 | 21.03% | 770s | 15,050 | $0.17 | $0.39 |
| 29th | MiMo-V2.5-Pro | 26.91 | 13.00 | 51.69% | 478s | 29,443 | $0.71 | $0.86 |
| 30th | Qwen3.7-Plus | 26.19 | 16.23 | 38.03% | 314s | 8,692 | $0.28 | $1.14 |
| 31st | Step-3.7-Flash | 26.18 | 13.74 | 47.52% | 258s | 44,912 | $1.46 | $1.16 |
| 32nd | Qwen3.5-27B | 24.96 | 17.98 | 27.96% | 451s | 26,391 | $0.51 | $0.69 |
| 33rd | Gemini 3.1 Flash Lite (high) | 23.78 | 14.30 | 39.87% | 82s | 28,351 | $1.17 | $1.48 |
| 34th | Claude Sonnet 5 | 23.64 | 10.86 | 54.06% | 109s | 5,962 | $1.65 | $9.86 |
| 35th | Qwen3.7-Max | 22.45 | 18.53 | 17.46% | 196s | 6,710 | $0.97 | $5.14 |
| 36th | ERNIE 5.1 | 20.49 | 15.49 | 24.4% | 457s | 24,093 | $1.73 | $2.57 |
| 37th | openPangu-2.0-Flash | 19.83 | 10.76 | 45.74% | 571s | 28,199 | $0.18 | $0.23 |
| 38th | LongCat-2.0 | 19.40 | 9.74 | 49.79% | 395s | 14,886 | $0.48 | $1.14 |
| 39th | DeepSeek V4 Flash | 19.39 | 13.37 | 31.05% | 63s | 5,286 | $0.04 | $0.29 |
| 40th | Mistral Medium 3.5 | 18.44 | 13.87 | 24.78% | 194s | 27,854 | $5.77 | $7.39 |
| 41st | Doubao-Seed-2.1-pro | 17.49 | 9.74 | 44.31% | 305s | 6,009 | $0.72 | $4.29 |
| 42nd | GLM-5.2 | 14.80 | 10.43 | 29.53% | 23s | 953 | $0.11 | $4.00 |
| 43rd | Ling-2.6-1T | 13.05 | 8.39 | 35.71% | 229s | 4,770 | $0.31 | $2.29 |
| 44th | iFLYTEK Spark X2 | 12.39 | 6.21 | 49.88% | 458s | 11,285 | $0.09 | $0.29 |
| 45th | Gemini 3.1 Flash Lite | 11.79 | 10.28 | 12.81% | 9s | 773 | $0.03 | $1.48 |
| 46th | Mistral Medium 3.5 | 10.83 | 7.13 | 34.16% | 17s | 2,546 | $0.53 | $7.39 |
| 47th | Ling-2.6-flash | 10.07 | 4.85 | 51.84% | 25s | 4,997 | $0.04 | $0.30 |
Quick Start
Connect DeepSeek V4 Pro API without changing the OpenAI-style request shape. Use DeepSeek V4 Pro for asynchronous reasoning, long text analysis, and cost-controlled batches. This entry is text only, and interactive products should test latency before routing live conversations.
Get your API key
Create an APINEED key for DeepSeek V4 Pro API and keep it in an environment variable.
export API_NEED_API_KEY=sk-apineed-v1-...Make your first request
Use deepseek-v4-pro for DeepSeek V4 Pro API with the APINEED API. The request shape is compatible with OpenAI chat completions, so most SDKs only need a base URL change.
TypeScript SDKPythoncURLOpenAI SDK
import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.API_NEED_API_KEY,
baseURL: "https://apineed.com/v1"
});
const result = await client.chat.completions.create({
model: "deepseek/deepseek-v4-pro",
messages: [
{ role: "user", content: "Why is the sky blue?" }
]
});
console.log(result.choices[0].message.content);Enable streaming and fallbacks
Add stream: true when DeepSeek V4 Pro API should return server-sent events. APINEED keeps routing, provider health, and fallback handling behind the same endpoint.
curl https://apineed.com/v1/chat/completions \
-H "Authorization: Bearer $API_NEED_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek/deepseek-v4-pro",
"stream": true,
"messages": [{ "role": "user", "content": "Hello" }]
}'DeepSeek V4 Pro API Endpoint
DeepSeek V4 Pro API accepts chat conversations here with streaming or non-streaming text output; the supported controls are listed on this page.
POST
/v1/chat/completions- Authorization
- Bearer $API_NEED_API_KEY
- Content-Type
- application/json
- HTTP-Referer
- optional - your site URL, for rankings
- X-Title
- optional - your site name, for rankings
- Model
- deepseek/deepseek-v4-pro
Parameters
DeepSeek V4 Pro API parameters currently listed for deepseek/deepseek-v4-pro by the OpenRouter Models API. [1]
| Name | Type | Status | Description |
|---|---|---|---|
frequency_penalty | number | Supported | Penalizes repeated token frequency. |
include_reasoning | boolean | Supported | Includes reasoning content in the response when available. |
logit_bias | object | Supported | Adjusts the likelihood of selected tokens. |
logprobs | boolean | Supported | Returns token log probabilities. |
max_tokens | integer | Supported | Limits generated output tokens. |
min_p | number | Supported | Applies minimum-probability sampling. |
presence_penalty | number | Supported | Penalizes tokens already present in the output. |
reasoning | object | Supported | Controls reasoning behavior and token allocation. |
reasoning_effort | enum | Supported | Selects the model reasoning effort. |
repetition_penalty | number | Supported | Controls repetition across generated tokens. |
response_format | object | Supported | Requests a specific response format. |
seed | integer | Supported | Requests deterministic sampling when supported by the provider. |
stop | string or array | Supported | Stops generation at the supplied sequence. |
structured_outputs | boolean | Supported | Enables schema-constrained structured output. |
temperature | number | Supported | Controls sampling randomness. |
tool_choice | string or object | Supported | Controls which tool the model may call. |
tools | array | Supported | Defines tools available to the model. |
top_k | integer | Supported | Restricts sampling to the highest-probability tokens. |
top_logprobs | integer | Supported | Sets how many top-token log probabilities are returned. |
top_p | number | Supported | Controls nucleus sampling. |
DeepSeek V4 Pro API Q&A
Model-specific answers for teams comparing DeepSeek V4 Pro API on capability, benchmark evidence, integration, and cost.
What is DeepSeek V4 Pro API’s main trade-off?
It combines very low token rates and a #13 logic rank with a long 1,468-second measured completion time in the benchmark.
Does DeepSeek V4 Pro API accept images?
No. This catalog entry supports text input and text output only.
Why is it suited to batch work for DeepSeek V4 Pro API?
Low output cost and long context are useful for large asynchronous jobs where completion time is not user-facing.
How does its price compare with GPT-5.6 Luna for DeepSeek V4 Pro?
For DeepSeek V4 Pro API, deepSeek’s listed official input and output rates are lower than GPT-5.6 Luna’s, while GPT-5.6 Luna records a much shorter benchmark completion time.
Which endpoint serves DeepSeek V4 Pro API?
Call DeepSeek V4 Pro API through https://apineed.com/v1/chat/completions with deepseek-v4-pro as the model value. Existing OpenAI SDK clients usually need only the APINEED base URL and key.
How do I validate DeepSeek V4 Pro API before release?
Evaluate DeepSeek V4 Pro API on representative prompts, record quality, latency, and token use, then choose reasoning settings and fallbacks from those results.
How to Deploy the DeepSeek V4 Pro API on apineed.com
Deploy DeepSeek V4 Pro API through APINEED after validating its model-specific trade-offs above. The API key, credit, and endpoint flow stays consistent across the catalog.
Add credits for DeepSeek V4 Pro API
Add credits on APINEED before deploying the DeepSeek V4 Pro API. Pay-as-you-go billing lets usage start small and scale with production traffic.
Get your API key
Create an APINEED API key for the DeepSeek V4 Pro API. The same key can call DeepSeek V4 Pro and other AI APIs through apineed.com.
Set the model slug
Use deepseek-v4-pro as the model value when you deploy the DeepSeek V4 Pro API. Keep the APINEED base URL at https://apineed.com/v1.
Send a request to DeepSeek V4 Pro API
Send chat completions or responses to the DeepSeek V4 Pro API from Claude Code, Codex, or any custom agent. The request keeps its OpenAI-compatible shape.