DeepSeek V4.1 Flash

deepseek/deepseek-v4.1-flash
NewFeatured
VisionToolsJSONReasoning
by DeepSeek · 2026-09-10

DeepSeek-V4.1-Flash is the smallest model in DeepSeek's new architecture family, released September 10, 2026, with native multimodal visual understanding — the production successor to both V4 Flash and the V4 Flash Vision experiment. The new architecture targets a higher capability ceiling, faster inference and higher throughput, and DeepSeek reports V4.1 Flash comprehensively surpasses V4 Pro on performance, cost, speed and total task time. It accepts text and images with text output, serves a 1M-token context window with up to 384K output tokens, and supports thinking (default, with selectable effort low / high / max) and non-thinking modes across the Chat Completions, Responses and Anthropic-compatible APIs, along with JSON output, tool calls and chat prefix completion; FIM works in non-thinking mode only. Model weights are open on Hugging Face (deepseek-ai/DeepSeek-V4.1-Flash). Official benchmarks at release: 90.6 on Terminal-Bench 2.1, 74.2 on DeepSWE v1.1, 65.4 on NL2Repo-Bench, 90.9 on GPQA Diamond, 63.9 on HLE with tools, and strong native-vision agent results (89.6 BabyVision, 78.9 Chartography with tools). Pricing was cut alongside the release: $0.15/M input (cache miss), $0.60/M output and $0.003/M on cache hits at off-peak rates, doubling during weekday peak hours (01:00-04:00 and 06:00-10:00 UTC); all other hours including weekends are off-peak.

ctx1M tokens
Max output384K
Inputtext + image
Outputtext
p50 TTFT406 ms
INPUT$0.15/ 1M tokens
OUTPUT$0.60/ 1M tokens
p50 TTFT406 ms7d
p95 TTFT1.46 s7d
TRAFFIC228.3Mtokens / 7d

DeepSeek V4.1 Flash is a DeepSeek model available through OrcaRouter, the OpenAI-compatible API gateway. It accepts text and image input, carries a context window of 1,048,576 tokens, and can…

What is DeepSeek V4.1 Flash?

Who is DeepSeek V4.1 Flash built for?

Why does the 1,048,576-token context window matter?

Code samples

Call from any SDK

OpenAI-compatible — keep the SDK you already use

  • OpenAI SDKhttps://api.orcarouter.ai/v1
  • Anthropic SDKhttps://api.orcarouter.ai
import os

from openai import OpenAI

client = OpenAI(
    base_url="https://api.orcarouter.ai/v1",
    api_key=os.environ["ORCAROUTER_API_KEY"],
)

response = client.chat.completions.create(
    model="deepseek/deepseek-v4.1-flash",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

Supported parameters

  • include_reasoning
  • logprobs
  • max_tokens
  • reasoning
  • reasoning_effort
  • response_format
  • stop
  • stream
  • stream_options
  • temperature
  • thinking
  • tool_choice
  • tools
  • top_logprobs
  • top_p
  • user_id

Pricing

Pricing
Input / 1M tokens · Off-peak$0.150
Output / 1M tokens · Off-peak$0.600
Cache read / 1M · Off-peak$0.0030
Peak hours01:00–04:00, 06:00–10:00 ×2 (UTC)
Input / 1M tokens · ×2$0.300
Output / 1M tokens · ×2$1.20
Cache read / 1M · ×2$0.0060
CurrencyUSD

Cost calculator

Tokens / month10MM
Input share70%%
Estimated / month $2.85 · With prompt caching $2.34

Estimate based on list price

Token & cost estimator

Input tokens: 20Cost per request: $0.000303

Estimate only — actual token counts depend on the provider's tokenizer.

Performance

p50 TTFT
406 ms
Output speed
215 tok/s
p95 TTFT
1.46 s
Error rate
0.07%

Public benchmarks

Agents' Last Exam
31.8
Automation-Bench
54.8
BabyVision (with tools)
89.6
Chartography (with tools)
78.9
CyberGym
88.1
DeepSWE v1.1
74.2
ExploitGym
15.3
GPQA Diamond
90.9
HLE
36.8
HLE (with tools)
63.9
MathArena Apex
65.6
NL2Repo-Bench
65.4
ProgramBench
20.3
SEC-Bench Pro
62.8
Terminal-Bench 2.1
90.6
Terminal-Bench 3.0
30.0
Terminal-Bench 4.0
31.2
ZeroBench-main (with tools)
49.0
Source: api-docs.deepseek.com

Community buzz

What developers are saying this week

Hacker News13 mentions · 7dup 13 vs the previous week

How it compares

DeepSeek V4.1 FlashDeepSeek V4 ProDeepSeek V4 FlashDeepSeek V4 Flash Vision (Exp)
Input $/M$0.15$0.73$0.24$0.24
Output $/M$0.60$2.18$0.73$0.73
Context1.0M1.0M1.0M1.0M
Quality8/108/107/107/10
Compare side-by-sideCompare side-by-sideCompare side-by-sideCompare side-by-side

FAQ

How much does DeepSeek V4.1 Flash cost on OrcaRouter?
OrcaRouter bills DeepSeek V4.1 Flash at $0.15 per 1M input tokens and $0.60 per 1M output tokens, passed through at the provider rate with zero markup. There is no separate charge described for the context window itself: 1,048,576 tokens is a limit, not a fee. Input tokens include text, images, and conversation history you send; output tokens cover everything generated, up to 384,000 tokens.
How large are the context window and maximum output?
DeepSeek V4.1 Flash has a context window of 1,048,576 tokens and a maximum output of 384,000 tokens. The input window lets you submit whole documents, transcripts, or repository snapshots in one call, while the output ceiling supports long single-pass artefacts such as specifications, migration plans, or extended diffs.
What is DeepSeek V4.1 Flash best at?
It is built for long-input, long-output work: whole-document analysis, large-repository code comprehension, multimodal review of text plus images, and workflows that need substantial generated answers rather than short replies. Its 90.9 score on GPQA Diamond also points to solid graduate-level science and technical reasoning. Input supports both text and images; output is text only.
How does it compare with larger frontier models?
Frontier models may lead on the hardest reasoning benchmarks and typically charge more per token, while DeepSeek V4.1 Flash offers a 1,048,576-token context window and a 384,000-token output ceiling at $0.15 per 1M input and $0.60 per 1M output tokens. For long-context and high-volume work it is often the practical choice; for the hardest individual reasoning steps, routing those calls to a costlier model on OrcaRouter is a reasonable pattern.
How is data handled when I call this model?
OrcaRouter routes requests to DeepSeek's API and bills at the provider rate with zero markup. Retention, training-use, and related terms are governed by the provider's policies rather than a separate arrangement described here, so confirm the provider's current terms before sending sensitive material. Redact identifiers you do not need and keep prompts to the minimum the task requires; a smaller context also lowers input cost at $0.15 per 1M tokens.
How do I call it through an OpenAI-compatible API?
Set your client's base URL to https://api.orcarouter.ai/v1 and use the model id deepseek/deepseek-v4.1-flash with your OrcaRouter API key. Requests and responses follow the OpenAI chat completions schema, so existing SDKs, streaming handlers, and tool-call parsing work unchanged. Migration is usually a base URL and model string change plus a re-run of your evaluation set.
Can DeepSeek V4.1 Flash accept images?
Yes. Image input is supported through the standard multimodal content array, with text and image parts in the same user message. Image tokens count as input and are billed at $0.15 per 1M input tokens on OrcaRouter. Output remains text only, so the model describes or extracts from images rather than generating them.
Is a 90.9 on GPQA Diamond enough to trust it for my task?
It is a useful signal, not a guarantee. GPQA Diamond measures graduate-level science reasoning under a fixed prompt format, which is relevant for technical question answering but says little about long-context recall, tone, or extraction accuracy on your data. Build a small evaluation set from real inputs, run it against DeepSeek V4.1 Flash and any alternative model ids on OrcaRouter, and compare cost and quality before routing production traffic.

Embed this badge

DeepSeek: DeepSeek V4.1 Flash$0.15/M in406ms p50via OrcaRouter
HTML <a href="https://www.orcarouter.ai/models/deepseek/deepseek-v4.1-flash" target="_blank"> <img src="https://www.orcarouter.ai/embed/deepseek/deepseek-v4.1-flash.svg" alt="DeepSeek: DeepSeek V4.1 Flash on OrcaRouter" /> </a>
Markdown [![DeepSeek: DeepSeek V4.1 Flash](https://www.orcarouter.ai/embed/deepseek/deepseek-v4.1-flash.svg)](https://www.orcarouter.ai/models/deepseek/deepseek-v4.1-flash)

Model card as data

GET /api/public/models/deepseek/deepseek-v4.1-flashOpen
Machine-readable:/llms.txt/llms-full.txt