DeepSeek V4.1 Flash
Use DeepSeek V4.1 Flash (deepseek-v4.1-flash-1017) on the GreenPT API with EU hosting, a 1M-token context, configurable reasoning, and discounted prompt caching.
Use deepseek-v4.1-flash-1017 for chat, reasoning, and coding through GreenPT's
OpenAI-compatible API. It is the 17 October 2026 snapshot of DeepSeek V4.1 Flash,
runs on EU infrastructure and has a 1M-token context window. It reads images,
calls tools, and returns JSON with response_format: { "type": "json_object" }.
The earlier deepseek-v4.1-flash ID stays available at its own price, and is the
DeepSeek V4.1 Flash ID on the US endpoint.
Make a request
Set GREENPT_API_KEY to your GreenPT API key, then run:
curl https://api.greenpt.ai/v1/chat/completions \
-H "Authorization: Bearer $GREENPT_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-v4.1-flash-1017",
"messages": [
{ "role": "user", "content": "Explain how a database index speeds up a query." }
],
"reasoning_effort": "none",
"max_tokens": 512
}'The GreenPT model ID is deepseek-v4.1-flash-1017. Keep the full ID, including
the 1017 snapshot date, when configuring an OpenAI-compatible client. See Coding with GreenPT
for client setup and Streaming for streamed responses.
Pricing
| Model ID | Input | Output | Cached input |
|---|---|---|---|
deepseek-v4.1-flash-1017 | €0.36 | €1.45 | €0.04 |
deepseek-v4.1-flash | €0.22 | €1.10 | €0.011 |
Prices are in EUR per million tokens.
These are GreenPT customer rates. Cached input replaces the normal input rate only for tokens reported as cache hits. Cache reuse is automatic when a request repeats an earlier prefix; a hit is not guaranteed. See Prompt caching and Pricing.
Reasoning effort
Accepted values are none, minimal, low, medium, and high. The example
disables reasoning for a short answer. For a task that needs deliberation,
set reasoning_effort to low, medium, or high and increase max_tokens.
Reasoning tokens consume the output budget. A small budget can be exhausted
before the model returns visible answer text. Omit reasoning_effort to use
the model's default. See Reasoning effort.
Energy and hosting
GreenPT estimates energy from input, output, and cached-input token counts. Both V4.1 Flash IDs adopt the DeepSeek V4 Flash 0731 configuration:
| Setting | Value |
|---|---|
| Input energy | 4.7 Wh per million tokens |
| Output energy | 280 Wh per million tokens |
| Cached-input energy | 0.2 Wh per million tokens |
| PUE multiplier | 1.3 |
| Carbon intensity | 0.193 kgCO2/kWh |
The token coefficients are estimates inherited from 0731, not a separate measurement of V4.1 Flash. They represent IT energy before the PUE multiplier. See Carbon calculations for the method and Processing locations for hosting details.
Existing integrations
Adding the 1017 snapshot does not change your selected model. The
deepseek-v4.1-flash and deepseek-v4-flash-0731 IDs remain separate catalogue
entries with their own pricing. Change the model ID explicitly when you want to
use the snapshot.