Skip to content

DeepSeek · Chat

DeepSeek V4 Flash 0731

DeepSeek 🇨🇳ChatDeepSeek familyOpen weights. One of its sources says the weights are publicly available.The maker currently plans to deprecate it within the next 180 days (3 Dec 2026). Plans can change.

Newer version: DeepSeek V4.1 Flash 1017. We compared release dates and version names in our catalogue.

From Baseten's listing:
A 284B-parameter mixture-of-experts model with 1M-token context that activates only 13B parameters at inference, making it a fast, low-cost sibling to V4 Pro.
llminfo IQ
132.2
Market median price
$0.175 blended / 1M
Max context
1.1M
Listings
205

Evidence and sources

🤗 deepseek-ai/DeepSeek-V4-Flash-0731 Model architecture, as its sources state it llminfo IQ 132.2, ranked among chat models. From 149 benchmarks. Plenty of data behind it.How llminfo IQ is computedCoding IQ 112.8, 95% interval 105–121. From 23 benchmark families across 6 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 (64.8). Cheaper per task than its benchmark results predict. From 32 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 resultsChatty (57 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate. By effort: default 57 · high 61.Median 68,384 output tokens per task at default effort.How the Yap score worksReleased 31 Jul 2026Maker HQ: CN (according to DeepSeek's own website)
All rankings: #9 on OpenRouter this month #16 on OpenRouter today #19 in Hacker News discussions this week #19 on OpenRouter this week #41 on NanoGPT recentlyAll rankings: #9 on OpenRouter this month #16 on OpenRouter today #19 in Hacker News discussions this week #19 on OpenRouter this week #41 on NanoGPT recentlySources: Alibaba Cloud Model Studio (DashScope) Multi-Region API, Baseten data, EmpirioLabs data, Mancer data, flex.ai API data, LiteLLM data, Models.dev data, llminfo researchPrimary data from Alibaba Cloud Model Studio (DashScope) Multi-Region API from Baseten: name, description from EmpirioLabs: primary mode from Mancer: architecture from flex.ai API: hugging face id from LiteLLM: deprecation date from Models.dev: weights availability from llminfo research: release date
Tags from its sources:instructtoolsreasoningopen-sourcemoelong-contextfunction-calling

Advantages and trade-offs

Compared with 10 chat models, picked mainly by popularity and IQ

Trade-offs

Deprecation scheduled

The maker plans to deprecate it on 3 Dec 2026, within the next 180 days.

We only make a claim when there's enough data on models of the same type. Updated 22 Nov, 04:34 UTC.Prices come from the model's maker where they sell it, otherwise their official partner, otherwise the market median. This comparison uses 9 maker's prices; 1 market median price.

Input & output

as stated by the model's providers
Text
Text

Capabilities

Supported operations and provider-stated features

ReasoningFunction callingStructured outputPrompt cachingWeb search

Listings

 Joingonkadeepseek-ai/DeepSeek-V4-Flash-0731Source: Models.dev data380K$0.0062$0.0185——
 Saggdeepseek-ai/DeepSeek-V4-Flash-0731Source: Models.dev data400K$0.0397$0.0794——
 NanoGPTdeepseek/deepseek-v4-flash-0731Sources: Models.dev data, NanoGPT1M$0.044cache read $0.0014$0.132——
 OpenRouterInference by Baidudeepseek/deepseek-v4-flash-0731 · fp8Source: OpenRouter data docs1M$0.044cache read $0.0014$0.132109747ms
 Hugging FaceInference by Waferdeepseek-ai/DeepSeek-V4-Flash-0731Source: Hugging Face Inference Providers data1M$0.0443$0.30——
 Kilodeepseek/deepseek-v4-flash-0731Sources: Kilo AI Gateway data, Models.dev data1M$0.0443cache read $0.035$0.30——
 Engydeepseek-v4-flash-0731Source: Models.dev data1M$0.045cache read $0.009$0.09——
 Nous Portaldeepseek/deepseek-v4-flash-0731:batch · batchSource: Nous Portal Inference Gateway data1M$0.048cache read $0.012$0.144——
 Eden AIInference by DeepInfradeepinfra/deepseek-ai/DeepSeek-V4-Flash-0731Source: Models.dev data1M$0.06cache read $0.015$0.18——
 Flowdotdeepseek-ai/DeepSeek-V4-Flash-0731Source: FlowDot data1M$0.06$0.18——

14 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

Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
DeepSeek V4.1 launch chart · settings unspecified
multimodal inputs explicitly ignored
BrowseComp5 results
Show 5 more resultsHide 5 results
Source: OpenRouter benchmark exports · Aug 2026
exa search · OpenRouter's search tool · up to 25 turns · high effort
parallel search · OpenRouter's search tool · up to 1 turn · high effort
parallel search · OpenRouter's search tool · up to 25 turns · high effort
parallel search · OpenRouter's search tool · up to 5 turns · high effort
perplexity search · OpenRouter's search tool · up to 25 turns · high effort
GAIA4 results
Source: GAIA Leaderboard data · Aug 2026
Difficulty level
Difficulty not reported
SWE-Bench ProClaude Code · temp 1.0 · top_p 0.95 · 256K · refined tasks;Claude 对照为官方值 · Reported by Qwen · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Source: OpenRouter benchmark exports
31 Jul 2026 version · 90-day results
Across 32 providers lowest 67.3% median 74.9% highest 79.8%OpenRouter's automatic routing 73.2%
Show results from 32 providers and OpenRouter's automatic routingHide results from 32 providers and OpenRouter's automatic routing
  • BaseTen
  • Together
  • NextBit
  • Cloudflare
  • SiliconFlow
  • GMICloud
  • Sail Research
  • CoreWeave
  • Cohere
  • Novita
  • AtlasCloud
  • Mancer 2
  • Alibaba
  • StreamLake
  • DeepInfra
  • Venice
  • Parasail
  • Baidu
  • Wafer
  • Reka
  • Morph
  • PrimeIntellect
  • Phala
  • Inceptron
  • Relace
  • OpenRouter's automatic routing
  • OpenInference
  • Makora
  • Fireworks
  • Nebius
  • DigitalOcean
  • Ambient
  • AkashML
Terminal-Bench3 related benchmarks · 5 results
Terminal-Bench 2.1
Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
DeepSeek V4.1 launch chart · settings unspecified
official model-card setting
Source: Hugging Face Official Benchmarks data · Reported by the model's developer. The source relaying it did not verify it. claim
Setting not stated
Terminal-Bench 3.0
Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
DeepSeek V4.1 launch chart · settings unspecified
Terminal-Bench 4.0
Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
DeepSeek V4.1 launch chart · settings unspecified

Science

GPQA Diamond37 results
Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
DeepSeek V4.1 launch chart · settings unspecified
Reported by Qwen · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Qwen3.8 Omni Flash 官方发布表
Source: OpenRouter benchmark exports
31 Jul 2026 version · 90-day results
Across 33 providers lowest 78.8% median 89.2% highest 90.7%OpenRouter's automatic routing 87.1%
Show results from 33 providers and OpenRouter's automatic routingHide results from 33 providers and OpenRouter's automatic routing
  • Alibaba
  • DeepInfra
  • Novita
  • GMICloud
  • Baidu
  • PrimeIntellect
  • NextBit
  • Wafer
  • Relace
  • AtlasCloud
  • AkashML
  • SiliconFlow
  • Parasail
  • Cohere
  • Ambient
  • Mancer 2
  • Phala
  • Cloudflare
  • Fireworks
  • Reka
  • Inceptron
  • DigitalOcean
  • Morph
  • StreamLake
  • CoreWeave
  • Venice
  • Nebius
  • Together
  • Makora
  • OpenRouter's automatic routing
  • Sail Research
  • BaseTen
  • Chutes
  • OpenInference

Reasoning

Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
DeepSeek V4.1 launch chart · settings unspecified
Source: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
Setting not stated
CoWorkBenchReported by Qwen · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
DeepSearchQA7 results
Show 7 more resultsHide 7 results
Source: OpenRouter benchmark exports · Aug 2026
exa search · OpenRouter's search tool · up to 25 turns · high effort
exa search · OpenRouter's search tool · up to 5 turns · high effort
parallel search · OpenRouter's search tool · up to 1 turn · high effort
parallel search · OpenRouter's search tool · up to 25 turns · high effort
parallel search · OpenRouter's search tool · up to 5 turns · high effort
perplexity search · OpenRouter's search tool · up to 25 turns · high effort
perplexity search · OpenRouter's search tool · up to 5 turns · high effort
Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
DeepSeek V4.1 launch chart · settings unspecified · text-only subset
Reported by Qwen · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
WideSearch7 results
Show 7 more resultsHide 7 results
Source: OpenRouter benchmark exports · Aug 2026
exa search · OpenRouter's search tool · up to 25 turns · high effort
exa search · OpenRouter's search tool · up to 5 turns · high effort
parallel search · OpenRouter's search tool · up to 1 turn · high effort
parallel search · OpenRouter's search tool · up to 25 turns · high effort
parallel search · OpenRouter's search tool · up to 5 turns · high effort
perplexity search · OpenRouter's search tool · up to 25 turns · high effort
perplexity search · OpenRouter's search tool · up to 5 turns · high effort

Long context

Coding

Codeforces RatingDeepSeek V4.1 launch chart · settings unspecified · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
CyberGym3 results
Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
DeepSeek V4.1 launch chart · settings unspecified
official model-card setting
Source: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
Setting not stated
DeepSWESource: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
official model-card setting
Source: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
Setting not stated
DSBench-Hard2 results
Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
official model-card setting
Source: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
Setting not stated
LiveCodeBench v6Qwen3.8 Omni Flash 官方发布表 · Reported by Qwen · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
NL2RepoSource: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
SEC-Bench ProDeepSeek V4.1 launch chart · settings unspecified · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
ValsSource: Vals AI

Multilingual

Benchmark Nusantara18 related benchmarks · 20 results
id-N terbuka (prompt open-py1, scorer open-py1, items 5ca9dfe11dc2, key eda1d1cddc75, public slice)
Source: Benchmark Nusantara (ibahasa) data docs · Aug 2026
repetition: 1, route deepseek/deepseek-v4-flash-0731, token cap 2400
id-B pilihan-ganda (prompt mcq-py1, scorer mcq-py1, items 3e5973f02ae8, key 4e94e1cb7f25, public slice)
Source: Benchmark Nusantara (ibahasa) data docs · Aug 2026
repetition: 1, route deepseek/deepseek-v4-flash-0731, token cap 2400
id-B pilihan-ganda (prompt mcq-py1, scorer mcq-py1, items 3e5973f02ae8, key 4e94e1cb7f25, withheld slice)
Source: Benchmark Nusantara (ibahasa) data docs · Aug 2026
repetition: 1, route deepseek/deepseek-v4-flash-0731, token cap 2400
id-B pilihan-ganda (prompt mcq-py1, scorer mcq-py1, items d3e8c01bc554, key 4e94e1cb7f25, public slice)
Source: Benchmark Nusantara (ibahasa) data docs · Aug 2026
repetition: 1, route deepseek/deepseek-v4-flash-0731, token cap 2400
id-N terbuka (prompt open-py1, scorer open-py1, items 5ca9dfe11dc2, key eda1d1cddc75, withheld slice)
Source: Benchmark Nusantara (ibahasa) data docs · Aug 2026
repetition: 1, route deepseek/deepseek-v4-flash-0731, token cap 2400
Show 14 more resultsHide 14 results
id-N terbuka (prompt open-py2, scorer open-py1, items 5ca9dfe11dc2, key eda1d1cddc75, public slice)
Source: Benchmark Nusantara (ibahasa) data docs · Aug 2026
repetition: 1, route deepseek/deepseek-v4-flash-0731, token cap 2400
id-N terbuka (prompt open-py2, scorer open-py1, items 5ca9dfe11dc2, key eda1d1cddc75, withheld slice)
Source: Benchmark Nusantara (ibahasa) data docs · Aug 2026
repetition: 1, route deepseek/deepseek-v4-flash-0731, token cap 2400
id-N terbuka (prompt open-py3, scorer open-py1, items 5ca9dfe11dc2, key eda1d1cddc75, public slice)
Source: Benchmark Nusantara (ibahasa) data docs · Aug 2026
repetition: 1, route deepseek/deepseek-v4-flash-0731, token cap 2400
id-N terbuka (prompt open-py3, scorer open-py1, items 5ca9dfe11dc2, key eda1d1cddc75, withheld slice)
Source: Benchmark Nusantara (ibahasa) data docs · Aug 2026
repetition: 1, route deepseek/deepseek-v4-flash-0731, token cap 2400
id-W terbuka (prompt open-py1, scorer open-py1, items 27af2c184856, key 82bbf3cc2f8a, public slice)
Source: Benchmark Nusantara (ibahasa) data docs · Aug 2026
repetition: 1, route deepseek/deepseek-v4-flash-0731, token cap 9600
id-W terbuka (prompt open-py1, scorer open-py1, items 27af2c184856, key 82bbf3cc2f8a, withheld slice)
Source: Benchmark Nusantara (ibahasa) data docs · Aug 2026
repetition: 1, route deepseek/deepseek-v4-flash-0731, token cap 9600
jv-A pilihan-ganda (prompt mcq-py1, scorer mcq-py1, items c0c83850858e, key 18e8ec378904, public slice)
Source: Benchmark Nusantara (ibahasa) data docs · Aug 2026
repetition: 1, route deepseek/deepseek-v4-flash-0731, token cap 2400
jv-B arti-kata (prompt arti-kata-v1, scorer kunci-alias-v1, items 4e682041d284, key 873b6f168502, public slice)
Source: Benchmark Nusantara (ibahasa) data docs
repetition: 1, route deepseek/deepseek-v4-flash-0731, token cap unrecorded
repetition: 2, route deepseek/deepseek-v4-flash-0731, token cap 4000
repetition: 3, route deepseek/deepseek-v4-flash-0731, token cap 4000

General

Human preference

Design Arena EloSource: Design Arena Leaderboards data · Nov 2026

Math

MathArena ApexDeepSeek V4.1 launch chart · settings unspecified · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.

Factuality

Instruction following

IFBenchQwen3.8 Omni Flash 官方发布表 · Reported by Qwen · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.

Knowledge

Other

DeepSWE v1.12 results
Reported by Qwen · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Claude Code / mini-SWE-agent 取较高值 · temp 1.0 · top_p 0.95 · 256K
Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
DeepSeek V4.1 launch chart · settings unspecified
NL2Repo-Bench2 results
Reported by Qwen · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Claude Code · 禁止访问目标仓库的 Bash 命令
Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
DeepSeek V4.1 launch chart · settings unspecified · launch-page table
SkillsBench · Vals6 related benchmarks · 6 results
With Skills · Accuracy
Reported by Vals AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Vals AI · updated 2026-09-27 · metadata version 1 · 87 tasks in trace charts · 3 trials/task/condition · per-task mean then task average · OpenHands-CLI@ · With Skills · system=deepseek/deepseek-v4-flash-0731 · provider=DeepSeek · reasoning_effort=high · max_output_tokens=384000
With Skills · Mean cost / test
Reported by Vals AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Vals AI · updated 2026-09-27 · metadata version 1 · 87 tasks in trace charts · 3 trials/task/condition · per-task mean then task average · OpenHands-CLI@ · With Skills · system=deepseek/deepseek-v4-flash-0731 · provider=DeepSeek · reasoning_effort=high · max_output_tokens=384000
With Skills · Mean duration / test
Reported by Vals AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Vals AI · updated 2026-09-27 · metadata version 1 · 87 tasks in trace charts · 3 trials/task/condition · per-task mean then task average · OpenHands-CLI@ · With Skills · system=deepseek/deepseek-v4-flash-0731 · provider=DeepSeek · reasoning_effort=high · max_output_tokens=384000
Without Skills · Accuracy
Reported by Vals AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Vals AI · updated 2026-09-27 · metadata version 1 · 87 tasks in trace charts · 3 trials/task/condition · per-task mean then task average · OpenHands-CLI@ · Without Skills · system=deepseek/deepseek-v4-flash-0731 · provider=DeepSeek · reasoning_effort=high · max_output_tokens=384000
Without Skills · Mean cost / test
Reported by Vals AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Vals AI · updated 2026-09-27 · metadata version 1 · 87 tasks in trace charts · 3 trials/task/condition · per-task mean then task average · OpenHands-CLI@ · Without Skills · system=deepseek/deepseek-v4-flash-0731 · provider=DeepSeek · reasoning_effort=high · max_output_tokens=384000
Without Skills · Mean duration / test
Reported by Vals AI · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Vals AI · updated 2026-09-27 · metadata version 1 · 87 tasks in trace charts · 3 trials/task/condition · per-task mean then task average · OpenHands-CLI@ · Without Skills · system=deepseek/deepseek-v4-flash-0731 · provider=DeepSeek · reasoning_effort=high · max_output_tokens=384000
Show 7 more resultsHide 7 results
Source: OpenRouter benchmark exports · Aug 2026
exa search · OpenRouter's search tool · up to 25 turns · high effort
exa search · OpenRouter's search tool · up to 5 turns · high effort
parallel search · OpenRouter's search tool · up to 1 turn · high effort
parallel search · OpenRouter's search tool · up to 25 turns · high effort
parallel search · OpenRouter's search tool · up to 5 turns · high effort
perplexity search · OpenRouter's search tool · up to 25 turns · high effort
perplexity search · OpenRouter's search tool · up to 5 turns · high effort

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. · Data by LLM Stats · SEA-HELM leaderboard, AI Singapore (AI Products team), National University of Singapore. Published aggregate scores; results licence not stated (SEA-HELM code is MIT). · Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · Benchmark Nusantara, ibahasa (M. Khoirul Huda), doi:10.5281/zenodo.21970463. Test results CC BY 4.0, attribution to ibahasa; aggregate results CSV only. (licence) · OpenRouter public benchmark leaderboard and search-lane exports, CC BY 4.0. Fractional source scores converted to percentages; immutable dated snapshots. (licence) · BullshitBench by Peter Gostev, MIT (licence) · EQ-Bench (Samuel J. Paech) — cite paech2023eqbench · MineBench, MIT (licence) · Hugging Face official benchmark leaderboards; licence varies by benchmark · Clinical Benchmarks (clinicalbenchmarks.ai), CC BY 4.0; fractions converted to percentages; source documents retain their own terms (licence)

Around the web

Also known as: @cf/deepseek-ai/deepseek-v4-flash-0731, alibaba/deepseek-v4-flash-0731, azure_ai/deepseek-v4-flash-0731, baseten/deepseek-ai/deepseek-v4-flash-0731, dashscope/deepseek-v4-flash-0731, databricks/databricks-deepseek-v4-flash-0731, deepinfra/deepseek-ai/deepseek-v4-flash-0731, deepinfra/deepseek-v4-flash-0731, deepseek v4 flash 0731, deepseek v4 flash 0731 fast, deepseek-ai/deepseek-v4-flash-0731, deepseek-ai/deepseek-v4-flash-0731-tee, deepseek-v4-flash-0731-fast, deepseek-v4-flash-0731-free, deepseek-v4-flash-0731@eu, deepseek/deepseek-v4-flash-0731, deepseek/deepseek-v4-flash-0731-fast, deepseek/deepseek-v4-flash-0731:batch, deepseek/deepseek-v4-flash-0731:thinking, deepseek/deepseek-v4-flash-20260731, empiriolabs/deepseek-v4-flash-0731, fireworks_ai/accounts/fireworks/models/deepseek-v4-flash-0731, fireworks_ai/deepseek-v4-flash-0731, flexai/deepseek-v4-flash-0731, flexai/deepseek-v4-flash-0731@eu, gmi/deepseek-ai/deepseek-v4-flash-0731, nebius/deepseek-ai/deepseek-v4-flash-0731, novita/deepseek-v4-flash-0731, novita/deepseek/deepseek-v4-flash-0731, parasail-deepseek-v4-flash-0731, perplexity/deepseek-v4-flash-0731, qwen/deepseek-v4-flash-0731, qwen_ai_platform/deepseek-v4-flash-0731, qwencloud/deepseek-v4-flash-0731, scaleway/deepseek-v4-flash-0731, sference/deepseek-ai/deepseek-v4-flash-0731, sference/deepseek-v4-flash-0731, streamlake/deepseek-v4-flash-0731, tencent/deepseek-v4-flash-0731, tensorx/deepseek-v4-flash-0731, tensorx/deepseek/deepseek-v4-flash-0731, together_ai/deepseek-ai/deepseek-v4-flash-0731, umans-deepseek-v4-flash-0731, wandb/deepseek-ai/deepseek-v4-flash-0731

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