DeepSeek V4.1 Flash 1017: which model id to choose

Short answer: TensorX if you need the EU, the deepseek/ 1017 row if cost comes first

LLMTR is a gateway that gives access to language models from different providers through one OpenAI-compatible API. 1017, the 17 October 2026 release of DeepSeek V4.1 Flash, is called on LLMTR with three model ids: tensorx/deepseek-v4.1-flash-1017-tensorx, deepseek/deepseek-v4.1-flash-1017, and deepseek/deepseek-flash, which follows DeepSeek's current Flash release. All three are called on the /v1/chat/completions endpoint at https://llmtr.com/v1 with the same API key from your llmtr.com account.

Two questions decide it. If you need a row that, per the provider's statement, processes requests inside the European Union and does not use inputs for model training, the option is the TensorX row: $0.50 input and $1.50 output per 1M tokens. Without that requirement, deepseek/deepseek-v4.1-flash-1017 serves the same release at $0.14 input and $0.42 output through the end of 23 October 2026 (TRT), and at $0.20 and $0.60 from 24 October 2026; its price is the same at every hour of the day.

• TensorX, model list · 2026-10-19
• DeepSeek, models and pricing · 2026-10-19

Which model id should you use for DeepSeek V4.1 Flash 1017?

The three ids serve the same model family but differ on four points: where the request is processed, whether the version stays fixed, whether the price changes with the hour of the day, and how reasoning and JSON behave. DeepSeek's pricing page gives today's version of deepseek-flash as DeepSeek-V4.1-Flash-1017; when DeepSeek releases a new Flash version, deepseek/deepseek-flash moves to it while the two 1017 ids stay on 1017.

The table shows the three rows with the values in the LLMTR catalog; prices are per 1M tokens, in the order input / cached input / output.

What does the same request cost on each id?

Without a cache hit, a 20,000-token prompt with a 2,000-token answer costs $0.013 on the TensorX row. On deepseek/deepseek-v4.1-flash-1017 the same request costs $0.00364 through the end of 23 October 2026 and $0.0052 from 24 October. On deepseek/deepseek-flash it is $0.0042 off-peak and $0.0084 on weekday peak hours. Because deepseek-flash reasons by default, it can produce more output tokens for the same question; reasoning tokens bill at the output price.

The prompt cache widens the gap in work that repeats a long system prompt or document on every request. If 15,000 tokens of the same 20,000-token prompt are read from the cache and the answer is again 2,000 tokens, the TensorX row costs $0.00745, deepseek/deepseek-v4.1-flash-1017 $0.001603 through the end of 23 October 2026 and $0.00229 from 24 October, and deepseek/deepseek-flash $0.001995 off-peak. To benefit from the cache, put the content that does not change at the start of the prompt.

To see the real cost, send the same prompt to all three ids. The Python example below sends a contract clause to the three ids, reads the token counts in usage, and prints the cost with cached input priced on its own. The example sends no reasoning_effort: the TensorX row does not accept that field.

OpenAI Python SDK example that sends the same prompt to the three DeepSeek V4.1 Flash 1017 ids through https://llmtr.com/v1/chat/completions and prints token usage and cost.

import os

from openai import OpenAI

client = OpenAI(base_url="https://llmtr.com/v1", api_key=os.environ["LLMTR_API_KEY"])

# Per 1M tokens: (input, cached input, output) on 19 October 2026.
# deepseek-v4.1-flash-1017: introductory price through 23 October 2026 (TRT),
# then 0.20 / 0.006 / 0.60. deepseek-flash: off-peak price; every item doubles
# on weekday peak hours (01:00-04:00 and 06:00-10:00 UTC).
# The TensorX row needs the account's data transfer consent (Settings) first.
PRICES = {
    "tensorx/deepseek-v4.1-flash-1017-tensorx": (0.50, 0.13, 1.50),
    "deepseek/deepseek-v4.1-flash-1017": (0.14, 0.0042, 0.42),
    "deepseek/deepseek-flash": (0.15, 0.003, 0.60),
}
with open("clause.txt", encoding="utf-8") as f:
    clause = f.read()
messages = [{"role": "user", "content": "Summarize this contract clause in three bullet points:\n\n" + clause}]

for model, (inp, cached, out) in PRICES.items():
    response = client.chat.completions.create(model=model, messages=messages, max_tokens=4096)
    usage = response.usage
    details = usage.prompt_tokens_details
    hit = (details.cached_tokens or 0) if details else 0
    cost = ((usage.prompt_tokens - hit) * inp + hit * cached + usage.completion_tokens * out) / 1_000_000
    print(model, usage.prompt_tokens, usage.completion_tokens, f"${cost:.6f}")

Version pinning, reasoning and JSON

If you tuned your prompts and thresholds to a specific version, use a fixed id: the two 1017 ids do not move to DeepSeek's next Flash version. deepseek/deepseek-flash always serves the current version, and its price doubles on weekdays between 01:00-04:00 and 06:00-10:00 UTC, which is 04:00-07:00 and 09:00-13:00 Türkiye time; weekends and Chinese public holidays are billed at the normal price all day.

Reasoning differs across the three ids. On deepseek/deepseek-v4.1-flash-1017 reasoning starts off; reasoning_effort: high turns it on, none keeps it off, and no other level is accepted. The TensorX row rejects a request that sends reasoning_effort with 400. deepseek/deepseek-flash reasons by default; even for short answers, set max_tokens above the answer you expect.

If you need JSON that follows a schema, the option is deepseek/deepseek-v4.1-flash-1017: it accepts json_schema inside response_format. deepseek-flash lists JSON mode; the TensorX row's card lists no JSON, so validate the output on your side. On the two deepseek/ ids you can send an image as a base64 data URL or as an https address.

• DeepSeek, models and pricing · 2026-10-19

Which one when?

The choices below follow from the differences in price, version and behaviour between the three ids.

• Requests must be processed inside the European Union per the provider's statement, and inputs must not be used for model training: tensorx/deepseek-v4.1-flash-1017-tensorx. First give consent for TensorX in the Data transfer consents card under Settings; without it the request returns 403 provider_consent_required.
• The version must stay fixed, the price must not change with the hour of the day, and you need JSON that follows a schema: deepseek/deepseek-v4.1-flash-1017.
• You want DeepSeek's current Flash release at all times and up to 393,216 output tokens in one response: deepseek/deepseek-flash. Scheduling batch jobs outside peak hours halves the cost.
• If you built your evaluation set on one version, choose one of the two 1017 ids so your results do not shift when the version changes.

Trying all three with the same key

Open an account on llmtr.com, top up your balance under Dashboard > Billing and create a key under Dashboard > API Keys. Set the base URL in your OpenAI-compatible SDK to https://llmtr.com/v1. The same key works on all three ids; the TensorX row only needs the account-level consent, and that consent covers every key on the account.

Each request is charged to your balance at the catalog price in force when it is processed. The price change on deepseek/deepseek-v4.1-flash-1017 happens on its own at 24 October 2026 00:00 (TRT); you do not need to change anything. Setting a spending limit on your API key caps the cost during the trial.

Comparing the DeepSeek V4.1 Flash 1017 ids

Send the same prompt to the three ids with one LLMTR key, compare token counts and cost, and choose by what you need.

1. Key and consent. Top up your balance on llmtr.com, create an API key and give data transfer consent for TensorX under Settings.
2. Same prompt. Send your own prompts with the three model ids to the https://llmtr.com/v1/chat/completions endpoint.
3. Choice. Compare the answers, token counts and cost, then choose the model id.

Frequently asked questions

Are the TensorX row and the deepseek/ 1017 row the same model?

Both serve the V4.1 Flash 1017 release DeepSeek published on 17 October 2026 and stay on that release. Price, where the request is processed, reasoning control and JSON support differ.

When does the price of deepseek/deepseek-v4.1-flash-1017 change?

$0.14 / $0.0042 / $0.42 through the end of 23 October 2026 (TRT), then $0.20 / $0.006 / $0.60 from 24 October 2026 00:00. The change happens on its own and there is no peak-hour multiplier.

Why does the TensorX model return 403?

TensorX models cannot be called until your account gives data transfer consent. Give it in the Data transfer consents card under Settings; the consent applies to every API key on the account.

Related posts

• DeepSeek GDPR: DeepSeek V4 Through an EU-Hosted API
• TensorX API: EU-Hosted Models Through One API

• TensorX DeepSeek V4.1 Flash 1017 model
• DeepSeek V4.1 Flash 1017 model
• DeepSeek V4.1 Flash model
• TensorX documentation
• DeepSeek peak-hour pricing