DeepSeek V4 Pro 0813 is the production release of DeepSeek V4 Pro on Novita AI. Novita now lists the model ID as deepseek/deepseek-v4-pro-0813, with 1,048,576 context tokens, 393,216 max output tokens, and OpenAI-compatible access through https://api.novita.ai/openai. The model page also shows serverless delivery, function calling, structured outputs, and reasoning support.
What Changed in DeepSeek V4 Pro 0813
This release is not just a rename. DeepSeek’s changelog marks the V4 family as officially released on August 13, 2026, and says pricing adjustments take effect on August 16, 2026, UTC. The Hugging Face model card for DeepSeek-V4-Pro-0813 also identifies it as the official release and lists the license as MIT.
The key release points are:
| Field | Value |
|---|---|
| Model ID | deepseek/deepseek-v4-pro-0813 |
| Release status | GA / official release |
| Context window | 1,048,576 tokens |
| Max output | 393,216 tokens |
| Input | Text |
| Output | Text |
| License | MIT |
DeepSeek V4 Pro 0813 DSpark Decoding Update
The official model card says 0813 is the official release of DeepSeek V4 Pro and that it adds a DSpark speculative decoding module to the preview structure. In plain terms, that means the serving path is built to keep output aligned while improving decode efficiency. For teams that already use DeepSeek V4 Pro for agentic coding or long-context reasoning, the important point is simple: this is the version meant for production use, not the earlier preview guide.
If you want the implementation angle, keep the DeepSeek V4 Pro long-context guide for request shaping and the DeepSeek V4 Pro vs Flash guide for routing decisions.
Current Novita AI Pricing
As of this check, Novita’s model page shows:
| Field | Value |
|---|---|
| Input price | $1.32 / M tokens |
| Cache read price | $0.35 / M tokens |
| Output price | $3.96 / M tokens |
| Endpoint family | chat/completions, responses, anthropic |
That is the pricing to use for planning and rollout checks on Novita AI now. Do not reuse the older preview-era pricing from the earlier DeepSeek V4 Pro articles.
Quick Start
from openai import OpenAI
client = OpenAI(
api_key="YOUR_NOVITA_API_KEY",
base_url="https://api.novita.ai/openai/v1",
)
response = client.chat.completions.create(
model="deepseek/deepseek-v4-pro-0813",
messages=[
{"role": "user", "content": "Summarize the key GA changes in DeepSeek V4 Pro 0813."}
],
temperature=0.2,
)
print(response.choices[0].message.content)When To Use It
Use DeepSeek V4 Pro 0813 when you want the current GA model on Novita AI for coding, planning, and long-context work. If you are still deciding between Pro and Flash, start with Flash for baseline traffic and reserve Pro 0813 for harder prompts.
FAQ
Is DeepSeek V4 Pro 0813 the same as the earlier preview release?
No. 0813 is the GA release on Novita AI and the earlier preview-era guidance should not be treated as current pricing or release status.
What is the model ID on Novita AI?
deepseek/deepseek-v4-pro-0813.
Where should I start if I want routing guidance?
Use Flash for baseline traffic and Pro 0813 for harder prompts, long-context work, and higher-failure-cost tasks.
Verified sources:
