V3
deepseek.v3-v1:0Pricing
List prices at Amazon Bedrock, in US dollars per million tokens.
| Tier | Input / 1M | Output / 1M | |
|---|---|---|---|
| Standard | $0.58 | $1.68 |
What it costs a month
At Amazon Bedrock list prices, 30 days.
- Chat assistant5,000 requests a day, 2K in / 0.5K out$300
- RAG over documents2,000 requests a day, 8K in / 0.6K out$339
- Coding agent1,000 requests a day, 40K in / 2K out$797
- Overnight batch20,000 requests a day, 4K in / 0.4K out$1,795
Where to run it
2 hosts serve V3. The comparison is on blended price, 3 parts input to 1 part output.
| Host | In / out per 1M | Model id | |
|---|---|---|---|
Amazon BedrockCloud deepseek.v3-v1:0 | $0.58 / $1.68 Retires 30 Mar 2027 | deepseek.v3-v1:0 | Retires 30 Mar 2027 |
ReplicateInference v3 | - | v3 |
Capabilities
Core
- Streaming
- Function calling
- JSON mode
- Structured outputs
Multimodal
- Vision
- Image input
- Audio
- Image generation
- Video
Advanced
- Reasoning
- Tool calling
- MCP
- Prompt caching
- Batch API
- Embeddings
- Takes in
- Text
- Produces
- Text
- Features
- Native structured output
- Open weights
- License: DeepSeek License (MIT code)Runtimes: SGLang, TGI, vLLM
Cheaper alternatives
Current models that cost less and keep most of what this one can do: tools, vision, reasoning, structured output, caching and enough context.
Same model, cheaper host
No host sells it for less than Amazon Bedrock.
Smaller DeepSeek models
No cheaper DeepSeek model keeps its key capabilities.
Other frontier labs
- 1% less
- 46% less
- 77% less
- 77% less
Open-weight picks
- 5% less
- 7% less
- 20% less
- 25% less
Model line
Newer and older versions, at each lab's own price.
Call it
Using Amazon Bedrock's API.
from litellm import completion
response = completion(
model="bedrock/deepseek.v3-v1:0",
messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)Quick answers
How much does V3 cost?
On Amazon Bedrock, V3 costs $0.58 per million input tokens and $1.68 per million output tokens.
Which APIs offer V3?
V3 runs on 2 APIs: Amazon Bedrock and Replicate.
What is V3's context window?
128K tokens, with up to 8K tokens in one reply.
Is V3 being retired?
Amazon Bedrock retires it on 30 Mar 2027. 1 host keeps serving it.
Is there a newer version of V3?
Yes. V3.2 is the newest model in this line.
What are cheaper alternatives to V3?
Models that cost less and keep most of what it can do: Gemini 3.5 Flash Lite on Google AI Studio (1% less), GPT-5.4 Nano on OpenAI (46% less) and Nemotron 3 Ultra 550B A55B on Perplexity (5% less). Savings are on a mix of 3 input tokens to 1 output token.
What is V3's API model id?
Use deepseek.v3-v1:0 with Amazon Bedrock. Other hosts use their own ids: v3 on Replicate.
Sources
Raw catalog entry for deepseek.v3-v1:0 on Amazon Bedrock
{
"deprecation_date": "2027-03-30",
"deprecation_source": "https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-deepseek-deepseek-v3-1.html",
"input_cost_per_token": 5.8e-7,
"litellm_provider": "bedrock_converse",
"max_input_tokens": 128000,
"max_output_tokens": 8000,
"mode": "chat",
"output_cost_per_token": 0.00000168,
"source": "https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-deepseek-deepseek-v3-1.html",
"supports_function_calling": true,
"supports_reasoning": true,
"supports_tool_choice": true,
"supports_native_structured_output": true
}