impulse

DeepSeek V3.2 Exp

Open weightsPreview

DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures.

API id DeepSeek-V3.2-ExpReleased 29 Sep 2025 (Sep 2025)
Input / 1M
$0.27
at GMI Cloud
Output / 1M
$0.41
at GMI Cloud
Context window
164K
tokens
Max output
66K
tokens per reply
Hosts
2
2 listings
Released
Sep 2025
29 Sep 2025

Pricing

List prices at GMI Cloud, in US dollars per million tokens.

TierInput / 1MOutput / 1M
Standard$0.27$0.41

What it costs a month

At GMI Cloud list prices, 30 days.

  • Chat assistant5,000 requests a day, 2K in / 0.5K out$112
  • RAG over documents2,000 requests a day, 8K in / 0.6K out$144
  • Coding agent1,000 requests a day, 40K in / 2K out$349
  • Overnight batch20,000 requests a day, 4K in / 0.4K out$746
Use your own numbers

Where to run it

2 hosts serve DeepSeek V3.2 Exp. The comparison is on blended price, 3 parts input to 1 part output.

HostIn / out per 1M
GMI CloudInference
DeepSeek-V3.2-Exp
$0.27 / $0.41
Novita AIInference
deepseek-v3.2-exp
$0.27 / $0.41

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
Open weights
License: DeepSeek License (MIT code)
Runs in about 1645 GB of GPU memory
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 GMI Cloud.

Smaller DeepSeek models

No cheaper DeepSeek model keeps its key capabilities.

Other frontier labs

  • Claude Haiku 5.5
    Anthropic · on Anthropic
    $0.10 / $0.50
    Has every key capability
    34% less
  • GPT-6 Luna
    OpenAI · on OpenAI
    $0.10 / $0.50
    Has every key capability
    34% less
  • GPT-4o Mini
    OpenAI · on OpenAI
    $0.15 / $0.60
    Has every key capability
    14% less

Open-weight picks

  • Mistral Small 2603
    Mistral AI · on Mistral AI
    $0.15 / $0.60
    Has every key capability
    14% less
  • Qwen3.8 27B
    Alibaba Qwen · on EmpirioLabs AI
    $0.17 / $0.50
    Has every key capability
    17% less
  • Nemotron 3.5 Super VL
    NVIDIA · on Kyma API
    $0.081 / $0.472
    Has every key capability
    41% less
  • Gemma 4 31B IT
    Google · on W&B Inference
    $0.10 / $0.34
    Has every key capability
    48% less

Call it

Using GMI Cloud's API.

Python
from litellm import completion

response = completion(
    model="gmi/DeepSeek-V3.2-Exp",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

Quick answers

How much does DeepSeek V3.2 Exp cost?

On GMI Cloud, DeepSeek V3.2 Exp costs $0.27 per million input tokens and $0.41 per million output tokens.

Which APIs offer DeepSeek V3.2 Exp?

DeepSeek V3.2 Exp runs on 2 APIs: GMI Cloud and Novita AI.

What is DeepSeek V3.2 Exp's context window?

164K tokens, with up to 66K tokens in one reply.

Is DeepSeek V3.2 Exp being retired?

No retirement date is scheduled for it on any API we track.

What are cheaper alternatives to DeepSeek V3.2 Exp?

Models that cost less and keep most of what it can do: Claude Haiku 5.5 on Anthropic (34% less), GPT-6 Luna on OpenAI (34% less) and Mistral Small 2603 on Mistral AI (14% less). Savings are on a mix of 3 input tokens to 1 output token.

When was DeepSeek V3.2 Exp released?

DeepSeek released it on 29 Sep 2025.

What is DeepSeek V3.2 Exp's API model id?

Use DeepSeek-V3.2-Exp with GMI Cloud. Other hosts use their own ids: deepseek-v3.2-exp on Novita AI.

Sources

Raw catalog entry for DeepSeek-V3.2-Exp on GMI Cloud
{
  "input_cost_per_token": 2.7e-7,
  "litellm_provider": "gmi",
  "max_input_tokens": 163840,
  "mode": "chat",
  "output_cost_per_token": 4.1e-7,
  "source": "https://ie-api.gmicloud.ai/v1/artifact/get_public_artifacts"
}