An LLM’s context window is its working memory: the maximum text it can consider at once—your prompt, earlier messages, and its reply. For example, with an 8,000-token window, a long chat may exceed it, so older messages get dropped.
DeepSeek V4.1 Flash
Save 50%Officialdeepseek/deepseek-v4.1-flash
DeepSeek V4.1 Flash — DeepSeek's September 2026 Flash model, with native image input and a 1M-token context.
Try DeepSeek V4.1 Flash here
Tune the request, see the current price ceiling, and inspect the real response without leaving this model page.
Sign in first to upload images to this vision model.
- Context window
- 1M
- Max output
- 384K
- Input modalities
- TextImage
- Output modalities
- Text
- Capabilities
- Vision · Reasoning · Tool calling · JSON mode · Coding · Prompt cache
- Upstream release
- 2026-09-10
- RouteMux name
- deepseek/deepseek-v4.1-flash
- Official ID
- deepseek-flash
What is DeepSeek V4.1 Flash
DeepSeek V4.1 Flash was released on September 10, 2026, and DeepSeek calls it the smallest model in its new architecture family. It is a 552B-parameter MoE built on a new Causal Encoder–Decoder design that activates 8B parameters when reading input and 16B when generating output. Against the previous generation, DeepSeek says its KV cache needs a quarter of the HBM and an eighth of the SSD storage. Image understanding is native, and the separate experimental vision model from the V4 era, V4 Flash Vision Exp, was retired together with V4 Flash on September 10.
DeepSeek's own model name for it is deepseek-flash, and on October 17 that name was upgraded in place to DeepSeek-V4.1-Flash-1017, with calls and pricing unchanged. DeepSeek reports Terminal-Bench 4.0 rising from 31.2 to 42.3 and HLE without tools from 36.8 to 42.6 between the two builds, along with shorter task runtimes and shorter outputs.
The context window is 1M tokens and output runs up to 384K tokens, with tool calling and JSON output. Thinking is on by default and can be switched off. DeepSeek implements three effort levels, low, high and max; RouteMux accepts low, medium, high and max, and DeepSeek treats medium as high. In thinking mode, temperature, presence_penalty and frequency_penalty have no effect, while top_p only applies in thinking mode, within 0.95–1.0 (lower values are raised to 0.95). It is callable over OpenAI Chat Completions, OpenAI Responses and Anthropic Messages, and RouteMux forwards requests that still name deepseek/deepseek-v4-flash or deepseek/deepseek-v4-flash-vision-exp to this model.
On RouteMux, V4.1 Flash costs $0.15 per 1M input tokens, $0.6 per 1M output tokens and $0.003 per 1M cached input tokens.
DeepSeek V4.1 Flash API pricing
Prices verified 2026-10-25
Cache read $0.003/M · Cache write —/M
Estimate cost
Key capabilities
- Native vision
- Image understanding is built into the model, so text and images go to the same model id with no separate vision variant.
- 8B in, 16B out
- A 552B-parameter MoE whose Causal Encoder–Decoder design activates different amounts of the model for input and output.
- Upgraded to 1017
- DeepSeek says the October 17 build cuts task runtime and output length, with calls and pricing unchanged.
- Thinking on or off
- Thinking is the default and can be turned off for direct answers. Effort levels are low, high and max.
- Three protocols
- OpenAI Chat, OpenAI Responses, and Anthropic Messages all route to this model.
Limited time10% extra Credits on your first top-up — up to 200 bonus Credits
Applies to your first top-up within 72 hours of signing up. Minimum $20, once per account.
Interface and protocols
Use any of these IDs to call this model via the API.
deepseek/deepseek-v4.1-flashdeepseek-flashReplace the ROUTEMUX_KEY placeholder with your API key. Create one →
from openai import OpenAI
client = OpenAI(
base_url="https://api.routemux.com/v1",
api_key="<ROUTEMUX_KEY>",
)
completion = client.chat.completions.create(
model="deepseek/deepseek-v4.1-flash",
messages=[{"role": "user", "content": "What is the meaning of life?"}],
)
print(completion.choices[0].message.content)Performance (last 30 days)
DeepSeek V4.1 Flash FAQ
›What is DeepSeek V4.1 Flash?
DeepSeek V4.1 Flash — DeepSeek's September 2026 Flash model, with native image input and a 1M-token context.
›How large is the context window of DeepSeek V4.1 Flash?
DeepSeek V4.1 Flash supports a context window of up to 1M tokens, with up to 384K output tokens per request.
›How much does the DeepSeek V4.1 Flash API cost?
Through RouteMux, DeepSeek V4.1 Flash costs $0.15 for input and $0.6 for output (USD per 1M tokens), below the official list price.
›How do I call DeepSeek V4.1 Flash via API?
DeepSeek V4.1 Flash is OpenAI-compatible: point your base URL at RouteMux and set the model field to deepseek/deepseek-v4.1-flash — no code changes needed.
›Which input and output modalities does DeepSeek V4.1 Flash support?
DeepSeek V4.1 Flash accepts text, image as input and produces text as output.
›When was DeepSeek V4.1 Flash released?
DeepSeek V4.1 Flash was released on September 10, 2026.
Sources
- DeepSeek — DeepSeek-V4.1-Flash release ↗
- DeepSeek — API change log ↗
- DeepSeek — Models & pricing ↗
- DeepSeek — Thinking mode ↗
Reviewed by the RouteMux team on 2026-10-24