Skip to content

DeepSeek · Chat

DeepSeek V4.1 Flash 1017

DeepSeek 🇨🇳ChatDeepSeek familyOpen weights. One of its sources says the weights are publicly available.
From Baseten's listing:
A 552B-parameter mixture-of-experts model with native vision and 1M-token context, activating just 8B parameters during prefill and 16B during decode for efficient reasoning, coding, and agentic workloads.
llminfo IQ
139.7
Maker's price
$0.525 blended / 1M
Max context
1.1M
Listings
323

Evidence and sources

🤗 deepseek-ai/DeepSeek-V4.1-Flash-1017 Model architecture, as its sources state it llminfo IQ 139.7, ranked among chat models. From 97 benchmarks. Plenty of data behind it.How llminfo IQ is computedCoding IQ 126.8, 95% interval 119–135. From 10 benchmark families across 5 sources. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.How Coding IQ is computedGood value (71.4). Cheaper per task than its benchmark results predict. From 30 benchmarks with measured task costs. Plenty of data behind it.How Value for money worksYap score: tokens used per coding task vs. models with the same resultsTypical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate. By effort: default 55 · max 62.Median 82,681 output tokens per task at default effort.How the Yap score worksReleased 17 Oct 2026Maker HQ: CN (according to DeepSeek's own website)
All rankings: #1 on OpenRouter today #1 on OpenRouter this month #1 on OpenRouter this week #4 on NanoGPT recentlySources: Arcee AI data, Cloudflare Workers AI, EmpirioLabs data, flex.ai API data, Models.dev data, Baseten data, llminfo editorial researchPrimary data from Arcee AI from Cloudflare Workers AI: architecture from EmpirioLabs: primary mode from flex.ai API: hugging face id from Models.dev: weights availability from Baseten: description from llminfo editorial research: release date
Tags from its sources:instructtoolsreasoningvisionopen-sourcemoemultimodallong-contextfunction-calling

Advantages and trade-offs

Compared with 10 chat models, picked mainly by popularity and IQ
We only make a claim when there's enough data on models of the same type. Updated 22 Nov, 06:07 UTC.Prices come from the model's maker where they sell it, otherwise their official partner, otherwise the market median. This comparison uses 8 maker's prices; 2 market median prices.

Input & output

as stated by the model's providers
Text
Images
Files
Text

Capabilities

Supported operations and provider-stated features

VisionReasoningFunction callingStructured outputPrompt cachingWeb search

Listings

 DeepSeekdeepseek-flashSources: DeepSeek data, LiteLLM data, Models.dev data1M$0.30cache read $0.006$1.20——
 NanoGPTInference by CoralbricksRegions: UNKdeepseek/deepseek-v4.1-flash-1017 · standardSource: NanoGPTNot published by the provider$0.0053cache read $0.0032$0.5985125.62,712ms
 OpenRouterInference by Particle AIdeepseek/deepseek-v4.1-flash-1017 · us/flexSource: OpenRouter data docs1M$0.03cache read $0.015$0.275145736ms
 Engydeepseek-v4.1-flash-1017Source: Models.dev data1M$0.03cache read $0.006$0.12——
 Nous Portaldeepseek/deepseek-v4.1-flash-1017:batch · batchSource: Nous Portal Inference Gateway data1M$0.035cache read $0.0029$0.149——
 Isoquantdeepseek-v4.1-flash-1017-fast · fastSource: Models.dev data1M$0.035cache read $0.005$0.37——
 Hugging FaceInference by Waferdeepseek-ai/DeepSeek-V4.1-Flash-1017Source: Hugging Face Inference Providers data1M$0.036$0.50——
 Ambientdeepseek/deepseek-v4.1-flash-1017 · fp4Sources: Ambient data, Models.dev data1M$0.045cache read $0.0025$0.25——
 Openinferencedeepseek-ai/DeepSeek-V4.1-Flash-1017 · fp4Source: OpenInference data1M$0.05cache read $0.002$0.80——
 Primedeepseek/deepseek-v4.1-flash-1017Source: Models.dev data1M$0.0568cache read $0.0031$0.70cache write $0.0568——

20 listings without a price are hidden.

Included in plans

Umans Code ProumansSubscription$20/moslows down past the allowancePlan page
Umans Code MaxumansSubscription$50/moslows down past the allowancePlan page

Benchmarks

Named effort levels run from lowest to highest. Select a result to see its source details.

Agentic

Agents' Last Exam · Pass RateDeepSeek 1017 first-party evaluation · reasoning effort, candidate sampling, exact run count, and unspecified harnesses not published · published pass rate · harness unspecified · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Arena AgentSource: LMArena data · Nov 2026
Source: OpenRouter benchmark exports
17 Oct 2026 version · 90-day results
Across 41 providers lowest 70.4% median 73.3% highest 77.8%OpenRouter's automatic routing 73.3%
Show results from 41 providers and OpenRouter's automatic routingHide results from 41 providers and OpenRouter's automatic routing
  • AtlasCloud
  • Pearl Research
  • ModelRun
  • BaseTen
  • CoreWeave
  • Alibaba
  • GMICloud
  • Fireworks
  • Parasail
  • Decart
  • Morph
  • Wafer
  • DeepInfra
  • Together
  • Baidu
  • Mistral
  • Near AI
  • Relace
  • Cloudflare
  • SiliconFlow
  • Ambient
  • Crusoe
  • OpenRouter's automatic routing
  • DekaLLM
  • Novita
  • PrimeIntellect
  • Nebius
  • Sail Research
  • Modal
  • StepFun
  • Particle AI
  • Phala
  • InferenceNet
  • OpenInference
  • Ionstream
  • Reka
  • StreamLake
  • Venice
  • River
  • Cohere
  • DigitalOcean
  • Makora
Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
DeepSeek 1017 first-party evaluation · reasoning effort, candidate sampling, exact run count, and unspecified harnesses not published · Claude Code harness · provider run, distinct from maintainer leaderboard
Source: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
Setting not stated

Science

GPQA Diamond45 results
Source: OpenRouter benchmark exports
17 Oct 2026 version · 90-day results
Across 43 providers lowest 83.9% median 90.6% highest 92.8%OpenRouter's automatic routing 90.1%
Show results from 43 providers and OpenRouter's automatic routingHide results from 43 providers and OpenRouter's automatic routing
  • Tenstorrent
  • Near AI
  • DeepInfra
  • DigitalOcean
  • SiliconFlow
  • DekaLLM
  • Baidu
  • Alibaba
  • Venice
  • GMICloud
  • Pearl Research
  • River
  • StreamLake
  • Ambient
  • StepFun
  • CoreWeave
  • PrimaLabs
  • InferenceNet
  • Nebius
  • Cohere
  • Modal
  • Fireworks
  • Morph
  • Decart
  • Wafer
  • Relace
  • Makora
  • Sail Research
  • AtlasCloud
  • Mistral
  • PrimeIntellect
  • Particle AI
  • Together
  • Parasail
  • BaseTen
  • OpenRouter's automatic routing
  • Cloudflare
  • Ionstream
  • Crusoe
  • ModelRun
  • Phala
  • OpenInference
  • Novita
  • Reka

Reasoning

AutomationBench v1.0.6DeepSeek 1017 first-party evaluation · reasoning effort, candidate sampling, exact run count, and unspecified harnesses not published · published main score · metric subtype and harness unspecified · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
BabyVision2 results
Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
DeepSeek 1017 first-party evaluation · reasoning effort, candidate sampling, exact run count, and unspecified harnesses not published · with tools · harness unspecified
Source: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
Setting not stated
Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
DeepSeek 1017 first-party evaluation · reasoning effort, candidate sampling, exact run count, and unspecified harnesses not published · no tools
Source: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
Setting not stated
ZEROBenchSource: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim

Coding

Arena ELO WebdevSource: LMArena data · Nov 2026
NL2RepoSource: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
ValsSource: Vals AI

Human preference

Design Arena EloSource: Design Arena Leaderboards data · Nov 2026

General

Math

Factuality

Other

Apex Bench v1.1DeepSeek 1017 first-party evaluation · reasoning effort, candidate sampling, exact run count, and unspecified harnesses not published · published main score · metric subtype and harness unspecified · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Chartography · With toolsDeepSeek 1017 first-party evaluation · reasoning effort, candidate sampling, exact run count, and unspecified harnesses not published · with tools · harness unspecified · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
DeepSWE v1.1DeepSeek 1017 first-party evaluation · reasoning effort, candidate sampling, exact run count, and unspecified harnesses not published · Mini-SWE-V2 harness · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
NL2Repo-BenchDeepSeek 1017 first-party evaluation · reasoning effort, candidate sampling, exact run count, and unspecified harnesses not published · published main score · harness unspecified · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
ProgramBench · Metric UnspecifiedDeepSeek 1017 first-party evaluation · reasoning effort, candidate sampling, exact run count, and unspecified harnesses not published · published main score · metric subtype unspecified · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
SuperCLUESource: SuperCLUE data
ValsSource: Vals AI
ZeroBench · MainDeepSeek 1017 first-party evaluation · reasoning effort, candidate sampling, exact run count, and unspecified harnesses not published · with tools · pass@5 · harness unspecified · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · Hugging Face official benchmark leaderboards; licence varies by benchmark · Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · LMArena, CC BY 4.0 (licence) · Data by LLM Stats · Clinical Benchmarks (clinicalbenchmarks.ai), CC BY 4.0; fractions converted to percentages; source documents retain their own terms (licence) · Deep20Bench edition 1.1, Patrick Heusser. Results CC BY 4.0; native question scores and fraction-to-percent success rates. (licence) · OpenRouter public benchmark leaderboard and search-lane exports, CC BY 4.0. Fractional source scores converted to percentages; immutable dated snapshots. (licence) · WeirdML v3 by Håvard Tveit Ihle (Norwegian Defence Research Establishment). API costs supported by Epoch AI, METR and NDRE.

Also known as: @cf/deepseek-ai/deepseek-v4.1-flash-1017, accounts/fireworks/models/deepseek-v4p1-flash-1017, alibaba/deepseek-v4.1-flash-1017, azure_ai/deepseek-v4.1-flash-1017, baseten/deepseek-ai/deepseek-v4.1-flash-1017, baseten/deepseek-ai/deepseek-v4.1-flash-1017-fast, cline-pass/deepseek-v4.1-flash-1017, cloudflare/@cf/deepseek-ai/deepseek-v4.1-flash-1017, cloudflare/deepseek-v4.1-flash-1017, databricks/databricks-deepseek-v4-1-flash-1017, deepinfra/deepseek-ai/deepseek-v4.1-flash-1017, deepinfra/deepseek-v4.1-flash-1017, deepseek v4.1 flash (1017), deepseek-ai/deepseek-v4.1-flash-1017, deepseek-ai/deepseek-v4.1-flash-1017-fast, deepseek-v4-1-flash-1017, deepseek-v4-1-flash-261017, deepseek-v4.1-flash-1017-fast, deepseek-v4.1-flash-1017@eu, deepseek-v4.1-flash-20261017, deepseek-v4.1-flash-20261017-max, deepseek-v4.1-flash:1017, deepseek/deepseek-v4.1-flash-1017, deepseek/deepseek-v4.1-flash-1017-fast, deepseek/deepseek-v4.1-flash-1017-thinking, deepseek/deepseek-v4.1-flash-1017:batch, deepseek/deepseek-v4.1-flash-1017:thinking, empiriolabs/deepseek-v4-1-flash-1017, fireworks/deepseek-v4.1-flash-1017, fireworks_ai/accounts/fireworks/models/deepseek-v4p1-flash-1017, fireworks_ai/accounts/fireworks/models/deepseek-v4p1-flash-1017-us, fireworks_ai/accounts/fireworks/routers/deepseek-v4p1-flash-1017-flex, fireworks_ai/accounts/fireworks/routers/deepseek-v4p1-flash-1017-us, flexai/deepseek-v4.1-flash-1017, gmi/deepseek-ai/deepseek-v4.1-flash-1017, hf:deepseek-ai/deepseek-v4.1-flash-1017, nebius/deepseek-ai/deepseek-v4.1-flash-1017, novita/deepseek-v4.1-flash-1017, parasail-deepseek-v41-flash-1017, qwen/deepseek-v4.1-flash-1017, runware/deepseek-v4.1-flash-1017, scaleway/deepseek-v4.1-flash-1017, scx-ai-gp/deepseek-v4.1-flash-1017, sference/deepseek-ai/deepseek-v4.1-flash-1017, streamlake/deepseek-v4.1-flash-1017, tencent/deepseek-v4.1-flash-1017, tensorx/deepseek/deepseek-v4.1-flash-1017, together-ai/deepseek-v4.1-flash-1017, together_ai/deepseek-ai/deepseek-v4.1-flash-1017, umans-deepseek-v4.1-flash-1017, wandb/deepseek-ai/deepseek-v4.1-flash-1017

Like this one. Tap or hover any dotted value, label or icon to see where it comes from and how it is measured. open details when tapped or hovered. How this data works