DeepSeek V4.1 Flash 1017

Oct 2026
DeepSeek V4.1 Flash 1017
Compare in
on
$0.26per task?
$0.30 / $1.20per 1M in / outcache $0.006
167 tok/sfirst token in 1 s?
1Mcontext

Rankingat max effort?

Strong at

S
Agents & work tasks#6 of 175Reasoning#29 of 334Long context#22 of 273
A
Coding#17 of 132Knowledge#52 of 318Vision#15 of 124

Alternatives?

ModelScoreCostPer taskWhat changes
CheaperGLM 5.3 FlashunstatedOpensame-55%$0.15 / $0.50Defaults to max · Self-hosted+4BetterMuse Spark 1.3max+22%+281%$1.60Only auto tool_choice is supported+5OpenGLM 5.3unstated+9%+310%$1.40 / $4.40No image input+4CheaperClaude Haiku 5.5unstated-4%-62%$0.10 / $0.50Remove sampling parameters+5CheaperMiMo v2.6 FlashunstatedOpen-16%-67%$0.14 / $0.28Pass back reasoning_content+3

Run it yourself

estimate
Fits on✗ 8 GB (phone, laptop)✗ 16 GB✗ 24 GB GPU✗ 48 GB GPU✗ 80 GB GPU
LicenceMITfree for commercial use

History

4 versions
30354045Oct 26Release dateVals Legal Research BenchV4 · JulV4.1 · SepV4.1 · Oct
30354045Oct 26Release dateVals Legal Research BenchV4 · JulV4.1 · Sep

Benchmarks

all categories
Chat & Knowledge
Humanity's Last Exam43.3#29
AA Omniscience Index4.7#52
CritPt (physics)24#20
AA Intelligence Index 4.3.245.1#17
AA Omniscience · Accuracy42.9#40
AA Omniscience · Hallucination Rate66.8#122
AA-LCR (long context) 1.182.7#22
GDP.pdf · AA all-pass17.2#34
GPQA Diamondlab90.9
HLE fulllab36.8
HLE text-onlylab39.1
MathArena Apexlab65.6
SEC-Bench Prolab62.8
ExploitGymlab15.3
AA Cyber Index successeslab41
Coding
Terminal-Bench 4.033.8#17
FrontierCode 1.143.7#8
SciCode55.9#25
SWE-bench Pro (edition unstated)lab56
DeepSWE (edition unstated)lab59.3
DeepSWE 1.1lab77
SciCode Verified pass@1 n=6lab77.9
Terminal-Bench 2.1lab90.6
Terminal-Bench 3.0lab30
ProgramBenchlab20.3
NL2Repo-Benchlab65.4
Codeforces Ratinglab3,471
Agents & Tool Use
AutomationBench-AA66.6#9
GDPval-AA (work tasks) 2.162.3#6
AA-Briefcase 1.1 · Elo1,570.4#12
CyberGymlab88.1
HLE with toolslab63.9
Automation-Benchlab54.8
Agents Last Examlab31.8
CyberGym-E2E AAlab23
Vision & Documents
MMMU-Pro (vision)80.6#15
Chartography with toolslab78.9
GDP.pdf AAlab12.8
BabyVision with toolslab89.6
ZeroBench-main with toolslab49
Dense200 bounding-box detectionlab3.3
ChartQA Prolab59.9

Watch out 5

Pass back reasoning_content · Chat Completions›

With thinking and tools, return all prior reasoning_content even on turns without tool calls or the API returns HTTP 400. Source ↗

Disable thinking mode first · Chat Completions›

Required and named tool choices return HTTP 400 in thinking mode; disable thinking first. Source ↗

Parameters have no effect›

In thinking mode, temperature, presence_penalty and frequency_penalty have no effect. Source ↗

Values below 0.95›

Flash clamps top_p below 0.95 to 0.95 while thinking and ignores it in non-thinking mode, using 1.0. Source ↗

Stateless API · Responses›

Responses does not support previous_response_id or conversation; preserve conversation history in your client. Source ↗

FAQ

How much does DeepSeek V4.1 Flash 1017 cost?›

$0.30 per 1M input tokens and $1.20 per 1M output tokens; cached input $0.006 per 1M; cheapest via DeepInfra at $0.06 / $0.18. A typical benchmark task costs about $0.26 at max effort.

Is DeepSeek V4.1 Flash 1017 free to use?›

There is no free API; the cheapest paid price is $0.06 per 1M input tokens. The weights are open, so you can run it yourself.

How good is DeepSeek V4.1 Flash 1017, and what is it best at?›

#29 of 334 in Chat & Knowledge, #17 of 132 in Coding, #9 of 133 in Agents & Tool Use, #58 of 195 in Writing & Language, #15 of 124 in Vision & Documents. Strongest at: Coding, Agents & work tasks, Reasoning, Long context, Knowledge, Vision.

Is DeepSeek V4.1 Flash 1017 better than DeepSeek V4.1 Flash?›

On Vals Legal Research Bench, DeepSeek V4.1 Flash 1017 scores 45.2 vs 41.4 — higher.

How can I use DeepSeek V4.1 Flash 1017?›

Through the API of Alibaba, AtlasCloud, Azure, Baidu, BaseTen and 17 more.

What is the context window of DeepSeek V4.1 Flash 1017?›

1M tokens, with up to 384K output tokens.

When was DeepSeek V4.1 Flash 1017 released? Is there a newer version?›

Released 2026-10-17. It is the current version in its line.

Is DeepSeek V4.1 Flash 1017 open source?›

Yes, the weights are open (MIT licence).

Can I run DeepSeek V4.1 Flash 1017 locally?›

The weights are open and can be self-hosted; hardware needs are not confirmed yet.

Can DeepSeek V4.1 Flash 1017 read images?›

It accepts text and images as input.

What is DeepSeek V4.1 Flash 1017?›

DeepSeek V4.1 Flash 1017 is an open-weights language model from DeepSeek, released Oct 2026, the successor to DeepSeek V4.1 Flash.

What are the best alternatives to DeepSeek V4.1 Flash 1017?›

GLM 5.3 Flash (cheaper), Muse Spark 1.3 (better), GLM 5.3 (open weights).