DeepSeek V3

deepseek/deepseek-chat
by DeepSeek · 2024-12-26

DeepSeek V3: 1M token context, large output, efficient Mixture-of-Experts text model

ctx1.05M tokens
Inputtext
Outputtext
p50 TTFT379 ms
INPUT$0.15/ 1M tokens
OUTPUT$0.29/ 1M tokens
p50 TTFT379 ms7d
p95 TTFT495 ms7d
TRAFFIC4.1Mtokens / 7d

Model details

What is DeepSeek V3 and who is it for?

DeepSeek V3 is a Mixture-of-Experts text model from DeepSeek, designed for tasks that require understanding and generating over very long contexts. Its 1,048,576-token context window allows processing of entire books, extensive codebases, or long conversation histories in a single pass. The model outputs up to 384,000 tokens, making it suitable for generating reports, articles, or complex structured data. It is intended for developers, researchers, and enterprises that need high-capacity language understanding without multimodal capabilities. The MoE architecture provides efficiency: only a fraction of parameters are used per generation, reducing latency and cost compared to dense models of similar scale. On OrcaRouter, it is billed at the provider’s rate with zero markup.

What input modalities does DeepSeek V3 support?

DeepSeek V3 supports text-only input. It does not accept images, audio, or video. All interactions must be via plain text prompts. This limitation means it is best for tasks that rely solely on language, such as document analysis, code completion, or text-based reasoning. For multimodal applications, consider models like GPT-4o or Llama 3.2 Vision.

How does DeepSeek V3 compare to other large-context models?

DeepSeek V3’s 1,048,576-token context window is among the largest available. For comparison, many models support 128k or 200k tokens. Its maximum output of 384,000 tokens is also notably high. The MoE architecture offers a cost advantage: it uses fewer compute per token, which translates to lower latency and per-token pricing. However, the model is text-only and may not match dense models on tasks requiring deep world knowledge or nuance. Benchmarks (not provided) show it competes strongly in reasoning and coding, but specific scores are not listed here.

Code samples

from openai import OpenAI

client = OpenAI(
    base_url="https://api.orcarouter.ai/v1",
    api_key="$ORCAROUTER_API_KEY",
)

response = client.chat.completions.create(
    model="deepseek/deepseek-chat",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

Pricing

Input / 1M tokens$0.147
Output / 1M tokens$0.295
Cache read / 1M$0.020
CurrencyUSD

Performance

p50 TTFT
379 ms
Output speed
74.5 tok/s
p95 TTFT
495 ms
Error rate
0.04%

Public benchmarks

Ranking Distribution9090 tournaments
2300
First
2676
Second
2636
Third
1478
Fourth
Category Performance
3DElo: 1166WebsiteElo: 1163Code CategoriesElo: 1158UI ComponentElo: 1149Data VisualizationElo: 1141Game DevElo: 1121SVGElo: 1034
3DTop 54%
1166Elo
50.7%Win
43.2sAvg
Top 54%Rank
WebsiteTop 62%
1163Elo
48.5%Win
79.6sAvg
Top 62%Rank
Code CategoriesTop 66%
1158Elo
48.5%Win
74.0sAvg
Top 66%Rank
UI ComponentTop 66%
1149Elo
52.8%Win
47.2sAvg
Top 66%Rank
Data VisualizationTop 69%
1141Elo
51.4%Win
47.6sAvg
Top 69%Rank
Game DevTop 73%
1121Elo
43.9%Win
70.9sAvg
Top 73%Rank
SVGTop 94%
1034Elo
38.8%Win
15.2sAvg
Top 94%Rank
Source: Design Arena

FAQ

What is the cost per token for DeepSeek V3 on OrcaRouter?
Input: $0.14 per 1 million tokens. Output: $0.28 per 1 million tokens. Billed at provider rate with zero markup via OrcaRouter.
How large is the context window for DeepSeek V3?
1,048,576 tokens (approximately 1 million). Maximum output is 384,000 tokens.
What are the main strengths of DeepSeek V3?
Extreme context window (1M tokens), high output limit, MoE efficiency leading to lower cost per token, strong performance in reasoning and coding tasks.
How does DeepSeek V3 compare to other large language models?
It offers a larger context window than GPT-4o (128k) or Claude 3.5 Sonnet (200k) and is often cheaper per token. However, it is text-only, while some alternatives support images. MoE architecture provides speed advantages.
Does OrcaRouter store my data when using DeepSeek V3?
OrcaRouter acts as a gateway and does not store prompts or completions. Data handling follows DeepSeek's policies; refer to OrcaRouter's privacy policy for details.
How do I call DeepSeek V3 via an OpenAI-compatible API?
Set base_url to https://api.orcarouter.ai/v1 and model to 'deepseek/deepseek-chat'. Use standard OpenAI client libraries.
What parameters can I adjust for DeepSeek V3?
All standard chat completion parameters: temperature, top_p, max_tokens (up to 384,000), stop, frequency_penalty, presence_penalty, etc.
Is DeepSeek V3 multimodal?
No, it supports only text input and output. No image, audio, or video processing.
Can I stream responses from DeepSeek V3 on OrcaRouter?
Yes, streaming is supported via the standard OpenAI streaming interface.
What is the typical latency for DeepSeek V3?
Latency varies with input/output length and load. MoE architecture generally results in faster generation per token compared to dense models of similar size. No specific figures are provided.

Embed this badge

DeepSeek: DeepSeek V3$0.15/M in379ms p50via OrcaRouter
HTML <a href="https://www.orcarouter.ai/models/deepseek/deepseek-chat" target="_blank"> <img src="https://www.orcarouter.ai/embed/deepseek/deepseek-chat.svg" alt="DeepSeek: DeepSeek V3 on OrcaRouter" /> </a>
Markdown [![DeepSeek: DeepSeek V3](https://www.orcarouter.ai/embed/deepseek/deepseek-chat.svg)](https://www.orcarouter.ai/models/deepseek/deepseek-chat)