DeepSeek V3.2 Exp API: toman pricing and code
deepseek/deepseek-v3.2-exp
toolsreasoningjsonYou are billed for the usage the request actually reported. Prices follow the market.Pricing and top-ups
DeepSeek V3.2 Exp example: a multi-step problem with reasoning
The example is picked from this model's own capabilities. Drop in your key and it runs as is.
from openai import OpenAI
client = OpenAI(base_url="https://api.uttapen.ir/v1", api_key="sk-up-…")
resp = client.chat.completions.create(
model="deepseek/deepseek-v3.2-exp",
messages=[{"role": "user", "content": (
"A shop has 3 warehouses. A ships 120 orders a day, B ships 85 and C ships 40. "
"If C closes and its load is split between A and B in proportion to their capacity, how many does each ship a day? "
"Work through it step by step and give just the two numbers at the end."
)}],
reasoning_effort="medium", # low | medium | high
)
print(resp.choices[0].message.content)
# reasoning tokens count as output tokens too:
print(resp.usage.completion_tokens_details)curl https://api.uttapen.ir/v1/chat/completions \
-H "Authorization: Bearer sk-up-…" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek/deepseek-v3.2-exp",
"messages": [{"role": "user", "content": "120 and 85 daily orders split across two warehouses; work through it step by step."}],
"reasoning": {"effort": "medium"}
}'import OpenAI from "openai";
const client = new OpenAI({ baseURL: "https://api.uttapen.ir/v1", apiKey: "sk-up-…" });
const resp = await client.chat.completions.create({
model: "deepseek/deepseek-v3.2-exp",
messages: [{ role: "user", content: "120 and 85 daily orders are split between two warehouses; work through it step by step and give just the two numbers at the end." }],
reasoning_effort: "medium",
});
console.log(resp.choices[0].message.content);
console.log(resp.usage.completion_tokens_details);What is DeepSeek V3.2 Exp good for?
DeepSeek V3.2 Exp comes from DeepSeek; in uttapen you reach it with the model id "deepseek/deepseek-v3.2-exp". It accepts up to 163,840 tokens of input per call, so a long document or several source files fit in a single request. A single response can run to 65,536 tokens. On price it sits in the "very cheap" band — cheaper than 82 and dearer than 324 of the other paid models in the catalogue.
What you get on top of text in, text out: it supports tool calling, so it can invoke your own functions with valid arguments; it returns schema-valid JSON through response_format, ready to hand to your code; it has a reasoning mode that pays off on multi-step problems, maths and debugging. All of it works through the standard parameters of the official OpenAI SDK — no custom client, no wrapper. Keep in mind that reasoning tokens are output tokens and do appear on the bill.
Input runs at 78,334 toman per 1M tokens and output at 118,951 — output costs 1.5× input, so trimming the answer saves more than trimming the prompt. You are always charged for the usage the request actually reported, never for the estimate, and a failed request costs nothing.
The closest alternative with the same capabilities from a different provider is Llama 3.1 70B Instruct: DeepSeek V3.2 Exp works out roughly 1.2× cheaper, and its context window is larger. Both run on the same key and the same code, so trying the other one is a single string change.
To make the figure concrete: 100,000 toman of credit buys roughly 391 thousand-word requests on DeepSeek V3.2 Exp, and every 1,000 toman is about 3,906 words of round trip. A job with one million input tokens and one million output tokens comes to 197,285 toman in total. Filling this model's 163,840-token window costs 12,834 toman on the input side alone, which is the real reason to keep conversation history short.
Its nearest relative in the catalogue is "deepseek/deepseek-v3.2", and choosing between those two is where most people hesitate. Both land at nearly the same price on a thousand-word request (256 and 252 toman).
DeepSeek has 17 models in our catalogue; the cheapest is DeepSeek V4 Flash Latest at 79 toman per thousand words and the dearest DeepSeek V4 Pro 0813 (batch) at 1,991. Among the less common parameters it accepts logit_bias, logprobs, min_p, repetition_penalty, top_k, top_logprobs — all through the standard request body. By context size the nearest option from another provider is Llama Guard 4 12B at 163,840 tokens.
Where it makes sense: high-volume work such as classification, tagging and bulk summarising, logic puzzles and code review, agents that reach out to APIs and databases, extracting data against a fixed schema.
DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism...
Three real jobs, priced on this model
Each figure is derived from the prices above and moves when they do.
| Job | Tokens | Cost |
|---|---|---|
| One chat turn with a medium history | 1,500 in + 400 out | 165 toman |
| Summarising a ten-page document | 4,000 in + 600 out | 385 toman |
| Classifying a thousand short rows | 120,000 in + 20,000 out | 11,779 toman |
Frequently asked
- How do I call DeepSeek V3.2 Exp from Iran?
- Sign up with your mobile number, top the wallet up in toman, create an API key, then in the official OpenAI SDK point base_url at https://api.uttapen.ir/v1 and set model to "deepseek/deepseek-v3.2-exp". Nothing else in your code changes.
- What does DeepSeek V3.2 Exp cost in toman?
- 78,334 toman per 1M input tokens and 118,951 toman per 1M output tokens; a 1,000-word request is around 256 toman. You pay for the usage that request actually reported, and a failed request costs nothing.
- How much input does DeepSeek V3.2 Exp take?
- Up to 163,840 tokens per request, roughly 123k words. A single answer can reach 65,536 tokens.
- Does DeepSeek V3.2 Exp support streaming and tool calling?
- Streaming (stream=true) works on every model here. This one supports tool calling in the standard OpenAI shape. Structured output through response_format works too.