Language modelLimited access

DeepSeek-V4-Flash

DeepSeek

Released: April 24, 2026
Data date: September 8, 2026
Predecessor: DeepSeek-V3.2
Successor: DeepSeek-V4.1-Flash

Benchmark-Meta-Score: Unavailable
Minimum coverage not met
Speed: 131.8 tokens/s
Output throughput
Price: $0.13 · $0.28
Input · Output / 1M tokens
Input: Text
Output: Text

What is distinctive about DeepSeek-V4-Flash is its throughput-oriented V4 design with 284 billion total parameters, 13 billion active parameters, and a 1,000,000-token context window. It activates fewer parameters per step than Pro.

DeepSeek discontinued the original V4-Flash endpoint in September 2026 and temporarily routed the name to DeepSeek-V4.1-Flash. The released V4-Flash checkpoint sits alongside DeepSeek-V3.2 and is not the current API target.

Specifications and access

Specification	Value and source
Weights	Open weightsSource
Access	API, LocalSource
Parameters	284BSource
Active parameters	13BSource
Context window	1,000,000 tokensSource
Notes	284B total, 13B active, 1M context, MIT-licensedSource
License	MITSource
Parameters	284 BSource
Active parameters	13 BSource
Architecture	MoE with CSA/HCA and mHCSource
Training tokens	>32TSource
Input	TextSource
Output	TextSource
Reasoning mode	non-think, think-high, think-maxSource
Tool calling	Yes (Native tool use is documented for V4 modes in official release documentation)Source
API model ID	deepseek-v4-flash (Pricing page updated alias: V4-Flash-0731)Source
Availability	First-party DeepSeek API retired 2026-09-10; deepseek-v4-flash requests route to DeepSeek-V4.1-Flash. (Released weights and third-party provider availability are separate from the first-party API lifecycle.) (retrieved 2026-09-23)Source

Benchmark meta-analysis

Overall and area scores

These Gradually indexes place selected published measurements on a 0 to 100 scale. They are not accuracy percentages and do not guarantee results in your application.

Retrieval dates of the sources used 2026-10-15 - 2026-10-19

Overall index

Not enough coverage for a score yet

provisional

Published configurations differ within this comparison cohort. Treat the index as provisional.

Panel-weight coverage: 7 %
Test families: 9 / 71
Sources: 2
Mathematical bounds with missing values: 3.9-96.5 / 100

Coding

Not enough coverage for a score yet

provisional

Published configurations differ within this comparison cohort. Treat the index as provisional.

Panel-weight coverage: 10 %
Test families: 2 / 15
Sources: 2
Mathematical bounds with missing values: 3.2-93.2 / 100

Reasoning

Not enough coverage for a score yet

provisional

Published configurations differ within this comparison cohort. Treat the index as provisional.

Panel-weight coverage: 36 %
Test families: 3 / 7
Sources: 1
Mathematical bounds with missing values: 19.9-84.1 / 100

Knowledge

Not enough coverage for a score yet

provisional

Published configurations differ within this comparison cohort. Treat the index as provisional.

Panel-weight coverage: 33 %
Test families: 1 / 3
Sources: 1
Mathematical bounds with missing values: 13.8-80.5 / 100

Agents

Not enough coverage for a score yet

provisional

Published configurations differ within this comparison cohort. Treat the index as provisional.

Panel-weight coverage: 25 %
Test families: 3 / 12
Sources: 1
Mathematical bounds with missing values: 17.6-92.6 / 100

Documented method · Included test families and measurements

Method version: benchmark-meta-score-v1

The index includes active measurements from reviewed primary sources when at least five models are compared in the same documented comparison slice across dataset, harness, agent, tools, and trial count. Each benchmark series and evaluator contributes one shared setup slice from the newest capture; when dates match, the wider slice is used. When one source and series include several explicitly documented reasoning tiers, the highest documented tier (none through max) is selected, never the best result value. The explicit source label “Non-reasoning” is treated as none. When selecting between multiple results, unknown or ambiguous tiers exclude the model from that series. Unambiguous single configurations do not require a tier ordering. The best individual result for each model is not selected. Equal raw values share the same rank percentile. Multiple series from the same evaluator within a family receive equal weight; evaluators within a family, families within an area, and areas in the overall index then receive equal weight. Unmeasured series remain visible in coverage and bounds. An overall score requires at least four areas, six test families, and two sources. An area score requires at least two test families. Every overall and area score also requires at least 50% of its panel weight.

Published model-specific parameters can differ. This is not a controlled experiment; comparisons with such differences are marked provisional.

Unmeasured portions are not entered as zero. They reduce coverage and contribute only to the displayed bounds. Scores with different coverage are therefore not equally complete.

The bounds hold observed percentiles fixed and allow missing contributions anywhere from 0 to 100. They are not confidence intervals or predictions of future recalculations.

The source count is not a count of independent experiments.

OECD handbook on composite indicators

All 9 test families and 9 of 9 raw measurements are listed here. The complete calculation for every model is available in the JSON export.

Test family: AA-Briefcase
Agents · Family index: 68.3 / 100
Nominal family weight in the overall panel: 0.6 % · Coverage of this test family: 100 %
- Artificial Analysis
  Published configurations differ within this cohort.
  Measurement:
  artificial-analysis-aa-briefcase
  Series weight within the test family:
  100 %
  Nominal contribution to the overall panel:
  0.6 %
  Raw value:
  1,258.64 Elo
  Metric:
  AA-Briefcase Elo
  Version:
  v1.1
  Participating models:
  83
  Rank percentile within the cohort:
  68.3 %
  Retrieved:
  2026-10-19
  Comparison cohort:
  artificial-analysis-aa-briefcase:v1.1:v4.3.2:reasoning:tokens=2000:suffix=Max
  Reasoning tier:
  Max
  Documented comparison setup
  Publication cohort:
  ["artificial-analysis-aa-briefcase","v1.1","aa-model-page-2026-10-19-8a022369738c"]
  Dataset type:
  artificial-analysis-evaluation
  Setup and model parameters
  Exact AA configuration: DeepSeek V4 Flash 0731 (Max) (release deepseek-v4-flash, row deepseek-v4-flash, UUID fe4c0848-e284-4e52-a79d-cdc28392f1a9, reasoning=true, source reasoning-token configuration budget=2000, suffix=Max). This configured value is not measured token consumption.. Reasoning: reasoning enabled. Reasoning effort: Max
  Exact AA configuration: DeepSeek V4 Flash 0731 (Max) (release deepseek-v4-flash, row deepseek-v4-flash, UUID fe4c0848-e284-4e52-a79d-cdc28392f1a9, reasoning=true, source reasoning-token configuration budget=2000, suffix=Max). This configured value is not measured token consumption.. Source model ID: artificial-analysis/fe4c0848-e284-4e52-a79d-cdc28392f1a9/deepseek-v4-flash/deepseek-v4-flash. Source model name: DeepSeek-V4-Flash. Provider reported by source: DeepSeek. Model snapshot: DeepSeek V4 Flash 0731 (Max). Evaluator: Artificial Analysis. Dataset type: artificial-analysis-evaluation. Reasoning: reasoning enabled. Reasoning effort: Max. Source ID: artificial-analysis. Evidence capture ID: aa-model-page-2026-10-19-8a022369738c. Source payload hash: 8a022369738cbe2a0e2f9bf486a2882865dbb0c6f1396dfec84059a802560e3c. Parser version: artificial-analysis-rsc-models-v2. Imported at: 2026-10-19
Test family: AA-LCR
Reasoning · Family index: 62.2 / 100
Nominal family weight in the overall panel: 1.02 % · Coverage of this test family: 100 %
- Artificial Analysis
  Published configurations differ within this cohort.
  Measurement:
  artificial-analysis-aa-lcr
  Series weight within the test family:
  100 %
  Nominal contribution to the overall panel:
  1.02 %
  Raw value:
  79.6666666666667 %
  Metric:
  Pass rate
  Version:
  v1.1
  Participating models:
  83
  Rank percentile within the cohort:
  62.2 %
  Retrieved:
  2026-10-19
  Comparison cohort:
  artificial-analysis-aa-lcr:v1.1:v4.3.2:reasoning:tokens=2000:suffix=Max
  Reasoning tier:
  Max
  Documented comparison setup
  Publication cohort:
  ["artificial-analysis-aa-lcr","v1.1","aa-model-page-2026-10-19-8a022369738c"]
  Dataset type:
  artificial-analysis-evaluation
  Setup and model parameters
  Exact AA configuration: DeepSeek V4 Flash 0731 (Max) (release deepseek-v4-flash, row deepseek-v4-flash, UUID fe4c0848-e284-4e52-a79d-cdc28392f1a9, reasoning=true, source reasoning-token configuration budget=2000, suffix=Max). This configured value is not measured token consumption.. Reasoning: reasoning enabled. Reasoning effort: Max
  Exact AA configuration: DeepSeek V4 Flash 0731 (Max) (release deepseek-v4-flash, row deepseek-v4-flash, UUID fe4c0848-e284-4e52-a79d-cdc28392f1a9, reasoning=true, source reasoning-token configuration budget=2000, suffix=Max). This configured value is not measured token consumption.. Source model ID: artificial-analysis/fe4c0848-e284-4e52-a79d-cdc28392f1a9/deepseek-v4-flash/deepseek-v4-flash. Source model name: DeepSeek-V4-Flash. Provider reported by source: DeepSeek. Model snapshot: DeepSeek V4 Flash 0731 (Max). Evaluator: Artificial Analysis. Dataset type: artificial-analysis-evaluation. Reasoning: reasoning enabled. Reasoning effort: Max. Score display: 79.7%. Source ID: artificial-analysis. Evidence capture ID: aa-model-page-2026-10-19-8a022369738c. Source payload hash: 8a022369738cbe2a0e2f9bf486a2882865dbb0c6f1396dfec84059a802560e3c. Parser version: artificial-analysis-rsc-models-v2. Imported at: 2026-10-19
Test family: AutomationBench-AA
Agents · Family index: 72.0 / 100
Nominal family weight in the overall panel: 0.6 % · Coverage of this test family: 100 %
- Artificial Analysis
  Published configurations differ within this cohort.
  Measurement:
  artificial-analysis-automationbench-aa
  Series weight within the test family:
  100 %
  Nominal contribution to the overall panel:
  0.6 %
  Raw value:
  53.96771855133674 %
  Metric:
  Objectives completed without guardrail violations
  Version:
  unspecified
  Participating models:
  83
  Rank percentile within the cohort:
  72.0 %
  Retrieved:
  2026-10-19
  Comparison cohort:
  artificial-analysis-automationbench-aa:unspecified:v4.3.2:reasoning:tokens=2000:suffix=Max
  Reasoning tier:
  Max
  Documented comparison setup
  Publication cohort:
  ["artificial-analysis-automationbench-aa","unspecified","aa-model-page-2026-10-19-8a022369738c"]
  Dataset type:
  artificial-analysis-evaluation
  Setup and model parameters
  Exact AA configuration: DeepSeek V4 Flash 0731 (Max) (release deepseek-v4-flash, row deepseek-v4-flash, UUID fe4c0848-e284-4e52-a79d-cdc28392f1a9, reasoning=true, source reasoning-token configuration budget=2000, suffix=Max). This configured value is not measured token consumption.. Reasoning: reasoning enabled. Reasoning effort: Max
  Exact AA configuration: DeepSeek V4 Flash 0731 (Max) (release deepseek-v4-flash, row deepseek-v4-flash, UUID fe4c0848-e284-4e52-a79d-cdc28392f1a9, reasoning=true, source reasoning-token configuration budget=2000, suffix=Max). This configured value is not measured token consumption.. Source model ID: artificial-analysis/fe4c0848-e284-4e52-a79d-cdc28392f1a9/deepseek-v4-flash/deepseek-v4-flash. Source model name: DeepSeek-V4-Flash. Provider reported by source: DeepSeek. Model snapshot: DeepSeek V4 Flash 0731 (Max). Evaluator: Artificial Analysis. Dataset type: artificial-analysis-evaluation. Reasoning: reasoning enabled. Reasoning effort: Max. Score display: 54.0%. Source ID: artificial-analysis. Evidence capture ID: aa-model-page-2026-10-19-8a022369738c. Source payload hash: 8a022369738cbe2a0e2f9bf486a2882865dbb0c6f1396dfec84059a802560e3c. Parser version: artificial-analysis-rsc-models-v2. Imported at: 2026-10-19
Test family: GDP.pdf
Reasoning · Family index: 47.0 / 100
Nominal family weight in the overall panel: 1.02 % · Coverage of this test family: 100 %
- Artificial Analysis
  Published configurations differ within this cohort.
  Measurement:
  artificial-analysis-gdp-pdf
  Series weight within the test family:
  100 %
  Nominal contribution to the overall panel:
  1.02 %
  Raw value:
  11 %
  Metric:
  All-pass rate
  Version:
  unspecified
  Participating models:
  83
  Rank percentile within the cohort:
  47.0 %
  Retrieved:
  2026-10-19
  Comparison cohort:
  artificial-analysis-gdp-pdf:unspecified:v4.3.2:reasoning:tokens=2000:suffix=Max
  Reasoning tier:
  Max
  Documented comparison setup
  Publication cohort:
  ["artificial-analysis-gdp-pdf","unspecified","aa-model-page-2026-10-19-8a022369738c"]
  Dataset type:
  artificial-analysis-evaluation
  Setup and model parameters
  Exact AA configuration: DeepSeek V4 Flash 0731 (Max) (release deepseek-v4-flash, row deepseek-v4-flash, UUID fe4c0848-e284-4e52-a79d-cdc28392f1a9, reasoning=true, source reasoning-token configuration budget=2000, suffix=Max). This configured value is not measured token consumption.. Reasoning: reasoning enabled. Reasoning effort: Max
  Exact AA configuration: DeepSeek V4 Flash 0731 (Max) (release deepseek-v4-flash, row deepseek-v4-flash, UUID fe4c0848-e284-4e52-a79d-cdc28392f1a9, reasoning=true, source reasoning-token configuration budget=2000, suffix=Max). This configured value is not measured token consumption.. Source model ID: artificial-analysis/fe4c0848-e284-4e52-a79d-cdc28392f1a9/deepseek-v4-flash/deepseek-v4-flash. Source model name: DeepSeek-V4-Flash. Provider reported by source: DeepSeek. Model snapshot: DeepSeek V4 Flash 0731 (Max). Evaluator: Artificial Analysis. Dataset type: artificial-analysis-evaluation. Reasoning: reasoning enabled. Reasoning effort: Max. Score display: 11.0%. Source ID: artificial-analysis. Evidence capture ID: aa-model-page-2026-10-19-8a022369738c. Source payload hash: 8a022369738cbe2a0e2f9bf486a2882865dbb0c6f1396dfec84059a802560e3c. Parser version: artificial-analysis-rsc-models-v2. Imported at: 2026-10-19
Test family: GDPval-AA
Agents · Family index: 70.7 / 100
Nominal family weight in the overall panel: 0.6 % · Coverage of this test family: 100 %
- Artificial Analysis
  Published configurations differ within this cohort.
  Measurement:
  artificial-analysis-gdpval-aa
  Series weight within the test family:
  100 %
  Nominal contribution to the overall panel:
  0.6 %
  Raw value:
  1,433.74 Elo
  Metric:
  GDPval-AA Elo
  Version:
  v2.1
  Participating models:
  83
  Rank percentile within the cohort:
  70.7 %
  Retrieved:
  2026-10-19
  Comparison cohort:
  artificial-analysis-gdpval-aa:v2.1:v4.3.2:reasoning:tokens=2000:suffix=Max
  Reasoning tier:
  Max
  Documented comparison setup
  Publication cohort:
  ["artificial-analysis-gdpval-aa","v2.1","aa-model-page-2026-10-19-8a022369738c"]
  Dataset type:
  artificial-analysis-evaluation
  Setup and model parameters
  Exact AA configuration: DeepSeek V4 Flash 0731 (Max) (release deepseek-v4-flash, row deepseek-v4-flash, UUID fe4c0848-e284-4e52-a79d-cdc28392f1a9, reasoning=true, source reasoning-token configuration budget=2000, suffix=Max). This configured value is not measured token consumption.. Reasoning: reasoning enabled. Reasoning effort: Max
  Exact AA configuration: DeepSeek V4 Flash 0731 (Max) (release deepseek-v4-flash, row deepseek-v4-flash, UUID fe4c0848-e284-4e52-a79d-cdc28392f1a9, reasoning=true, source reasoning-token configuration budget=2000, suffix=Max). This configured value is not measured token consumption.. Source model ID: artificial-analysis/fe4c0848-e284-4e52-a79d-cdc28392f1a9/deepseek-v4-flash/deepseek-v4-flash. Source model name: DeepSeek-V4-Flash. Provider reported by source: DeepSeek. Model snapshot: DeepSeek V4 Flash 0731 (Max). Evaluator: Artificial Analysis. Dataset type: artificial-analysis-evaluation. Reasoning: reasoning enabled. Reasoning effort: Max. Source ID: artificial-analysis. Evidence capture ID: aa-model-page-2026-10-19-8a022369738c. Source payload hash: 8a022369738cbe2a0e2f9bf486a2882865dbb0c6f1396dfec84059a802560e3c. Parser version: artificial-analysis-rsc-models-v2. Imported at: 2026-10-19
Test family: AA-Omniscience
Knowledge · Family index: 41.5 / 100
Nominal family weight in the overall panel: 2.38 % · Coverage of this test family: 100 %
- Artificial Analysis
  Published configurations differ within this cohort.
  Measurement:
  artificial-analysis-omniscience
  Series weight within the test family:
  100 %
  Nominal contribution to the overall panel:
  2.38 %
  Raw value:
  -14.283333333333333 points
  Metric:
  AA-Omniscience Index
  Version:
  unspecified
  Participating models:
  83
  Rank percentile within the cohort:
  41.5 %
  Retrieved:
  2026-10-19
  Comparison cohort:
  artificial-analysis-omniscience:unspecified:v4.3.2:reasoning:tokens=2000:suffix=Max
  Reasoning tier:
  Max
  Documented comparison setup
  Publication cohort:
  ["artificial-analysis-omniscience","unspecified","aa-model-page-2026-10-19-8a022369738c"]
  Dataset type:
  artificial-analysis-evaluation
  Setup and model parameters
  Exact AA configuration: DeepSeek V4 Flash 0731 (Max) (release deepseek-v4-flash, row deepseek-v4-flash, UUID fe4c0848-e284-4e52-a79d-cdc28392f1a9, reasoning=true, source reasoning-token configuration budget=2000, suffix=Max). This configured value is not measured token consumption.. Reasoning: reasoning enabled. Reasoning effort: Max
  Exact AA configuration: DeepSeek V4 Flash 0731 (Max) (release deepseek-v4-flash, row deepseek-v4-flash, UUID fe4c0848-e284-4e52-a79d-cdc28392f1a9, reasoning=true, source reasoning-token configuration budget=2000, suffix=Max). This configured value is not measured token consumption.. Source model ID: artificial-analysis/fe4c0848-e284-4e52-a79d-cdc28392f1a9/deepseek-v4-flash/deepseek-v4-flash. Source model name: DeepSeek-V4-Flash. Provider reported by source: DeepSeek. Model snapshot: DeepSeek V4 Flash 0731 (Max). Evaluator: Artificial Analysis. Dataset type: artificial-analysis-evaluation. Reasoning: reasoning enabled. Reasoning effort: Max. Source ID: artificial-analysis. Evidence capture ID: aa-model-page-2026-10-19-8a022369738c. Source payload hash: 8a022369738cbe2a0e2f9bf486a2882865dbb0c6f1396dfec84059a802560e3c. Parser version: artificial-analysis-rsc-models-v2. Imported at: 2026-10-19
Test family: Humanity’s Last Exam
Reasoning · Family index: 59.8 / 100
Nominal family weight in the overall panel: 1.02 % · Coverage of this test family: 50 %
- Artificial Analysis
  Published configurations differ within this cohort.
  Measurement:
  artificial-analysis-humanitys-last-exam
  Series weight within the test family:
  50 %
  Nominal contribution to the overall panel:
  0.51 %
  Raw value:
  38.5542168674699 %
  Metric:
  Text-only pass@1 accuracy
  Version:
  unspecified
  Participating models:
  83
  Rank percentile within the cohort:
  59.8 %
  Retrieved:
  2026-10-19
  Comparison cohort:
  artificial-analysis-humanitys-last-exam:unspecified:v4.3.2:reasoning:tokens=2000:suffix=Max
  Reasoning tier:
  Max
  Documented comparison setup
  Publication cohort:
  ["artificial-analysis-humanitys-last-exam","unspecified","aa-model-page-2026-10-19-8a022369738c"]
  Dataset type:
  artificial-analysis-evaluation
  Setup and model parameters
  Exact AA configuration: DeepSeek V4 Flash 0731 (Max) (release deepseek-v4-flash, row deepseek-v4-flash, UUID fe4c0848-e284-4e52-a79d-cdc28392f1a9, reasoning=true, source reasoning-token configuration budget=2000, suffix=Max). This configured value is not measured token consumption.. Reasoning: reasoning enabled. Reasoning effort: Max
  Exact AA configuration: DeepSeek V4 Flash 0731 (Max) (release deepseek-v4-flash, row deepseek-v4-flash, UUID fe4c0848-e284-4e52-a79d-cdc28392f1a9, reasoning=true, source reasoning-token configuration budget=2000, suffix=Max). This configured value is not measured token consumption.. Source model ID: artificial-analysis/fe4c0848-e284-4e52-a79d-cdc28392f1a9/deepseek-v4-flash/deepseek-v4-flash. Source model name: DeepSeek-V4-Flash. Provider reported by source: DeepSeek. Model snapshot: DeepSeek V4 Flash 0731 (Max). Evaluator: Artificial Analysis. Dataset type: artificial-analysis-evaluation. Reasoning: reasoning enabled. Reasoning effort: Max. Score display: 38.6%. Source ID: artificial-analysis. Evidence capture ID: aa-model-page-2026-10-19-8a022369738c. Source payload hash: 8a022369738cbe2a0e2f9bf486a2882865dbb0c6f1396dfec84059a802560e3c. Parser version: artificial-analysis-rsc-models-v2. Imported at: 2026-10-19
Test family: LMArena Code Arena WebDev
Coding · Family index: 13.9 / 100
Nominal family weight in the overall panel: 0.48 % · Coverage of this test family: 100 %
- LMArena
  Published configurations differ within this cohort.
  Measurement:
  lmarena-code-webdev
  Series weight within the test family:
  100 %
  Nominal contribution to the overall panel:
  0.48 %
  Raw value:
  1,430.6320866639408 rating
  Metric:
  WebDev Arena Bradley-Terry rating
  Version:
  webdev
  Participating models:
  37
  Rank percentile within the cohort:
  13.9 %
  Retrieved:
  2026-10-15
  Comparison cohort:
  lmarena-code-webdev:2026-10-13:overall
  Reasoning tier:
  high
  Documented comparison setup
  Publication cohort:
  lmarena-code-webdev:2026-10-13:overall
  Dataset type:
  public-crowdsourced-preferences
  Harness:
  Arena Code Arena WebDev
  Setup and model parameters
  Arena rating from 10,152 published pairwise votes.. Harness: Arena Code Arena WebDev. Reasoning effort: high
  Arena rating from 10,152 published pairwise votes.. Source model ID: lmarena/deepseek-v4-flash-high-preview. Source model name: DeepSeek-V4-Flash. Provider reported by source: DeepSeek. Model snapshot: deepseek-v4-flash-high-preview. Evaluator: LMArena. Dataset type: public-crowdsourced-preferences. Leaderboard date: 2026-10-13. Harness: Arena Code Arena WebDev. Reasoning effort: high. Source ID: lmarena. Evidence capture ID: lmarena-webdev-4589dee83b594891961ae5190c508a9288d2c2ae-28f8076e6c27. Source payload hash: 28f8076e6c27f1f89239f3d7243456357dfa175b078646a6bd859935b162cc9c. Parser version: lmarena-dataset-v1. Imported at: 2026-10-15
Test family: Terminal-Bench
Coding · Family index: 68.3 / 100
Nominal family weight in the overall panel: 0.48 % · Coverage of this test family: 50 %
- Artificial Analysis
  Published configurations differ within this cohort.
  Measurement:
  artificial-analysis-terminal-bench-4
  Series weight within the test family:
  50 %
  Nominal contribution to the overall panel:
  0.24 %
  Raw value:
  12.1212121212121 %
  Metric:
  Task success
  Version:
  4.0
  Participating models:
  83
  Rank percentile within the cohort:
  68.3 %
  Retrieved:
  2026-10-19
  Comparison cohort:
  artificial-analysis-terminal-bench-4:4.0:v4.3.2:reasoning:tokens=2000:suffix=Max
  Reasoning tier:
  Max
  Documented comparison setup
  Publication cohort:
  ["artificial-analysis-terminal-bench-4","4.0","aa-model-page-2026-10-19-8a022369738c"]
  Dataset type:
  artificial-analysis-evaluation
  Setup and model parameters
  Exact AA configuration: DeepSeek V4 Flash 0731 (Max) (release deepseek-v4-flash, row deepseek-v4-flash, UUID fe4c0848-e284-4e52-a79d-cdc28392f1a9, reasoning=true, source reasoning-token configuration budget=2000, suffix=Max). This configured value is not measured token consumption.. Reasoning: reasoning enabled. Reasoning effort: Max
  Exact AA configuration: DeepSeek V4 Flash 0731 (Max) (release deepseek-v4-flash, row deepseek-v4-flash, UUID fe4c0848-e284-4e52-a79d-cdc28392f1a9, reasoning=true, source reasoning-token configuration budget=2000, suffix=Max). This configured value is not measured token consumption.. Source model ID: artificial-analysis/fe4c0848-e284-4e52-a79d-cdc28392f1a9/deepseek-v4-flash/deepseek-v4-flash. Source model name: DeepSeek-V4-Flash. Provider reported by source: DeepSeek. Model snapshot: DeepSeek V4 Flash 0731 (Max). Evaluator: Artificial Analysis. Dataset type: artificial-analysis-evaluation. Reasoning: reasoning enabled. Reasoning effort: Max. Score display: 12.1%. Source ID: artificial-analysis. Evidence capture ID: aa-model-page-2026-10-19-8a022369738c. Source payload hash: 8a022369738cbe2a0e2f9bf486a2882865dbb0c6f1396dfec84059a802560e3c. Parser version: artificial-analysis-rsc-models-v2. Imported at: 2026-10-19

Benchmark sources at a glance

Each evaluator measures different capabilities under its own conditions. Test series, composite indices and their components stay separate. The detailed results below retain their scales, configurations and source dates.

Evaluator	Test series	Composite indices	Recorded active measurements
Artificial Analysis	8	1	11
LMArena	2	0	62

Coverage is not a quality ranking. Overlapping tests and multiple settings do not count as independent experiments. Vendor claims, historical and disputed results are excluded from this overview and remain labelled in the detailed evidence.

Published benchmarks

AlmanBench v0.1

Accepted cases

1,029 German tasks test conversion from standard German into a defined simplified-language specification. It measures the Alman specification, not general German writing quality.

1.GPT-5.595.43%

2.GPT-5.6 Sol95.14%

3.Claude Opus 594.95%

4.DeepSeek-V4-Flash93%

5.Kimi K390.96%

higher is betterAs of August 9, 2026Scale: 0% - 100%

Source: Alman Institut

gradually.ai

Methodology and context

Comparison cohort: almanbench-v0-1:official:overall
Metric: accepted cases
Scale: 0-100
Participants: 5
Methodology and context: 1,029 public tasks with one direct run per row. The target is the Alman language specification.. Source model ID: curated/deepseek-v4-flash. Source model name: DeepSeek-V4-Flash. Provider reported by source: DeepSeek. Evaluator: Alman Institut. Dataset type: public. Leaderboard date: 2026-08-05. Reasoning effort: max. Standard error: 0.795242. Source ID: alman-institut. Evidence capture ID: curated-alman-institut-2026-08-09-743e705129f5. Source payload hash: 743e705129f503f6fd3b52d59f38664050d06d8c02c2c26c94fc49f51f94e213. Parser version: curated-v1. Imported at: 2026-10-15

Source: Alman Institut

Taste-Bench

Overall

502 technical decision points from engineering and research workflows. Each choice is presented in both answer orders. The overall score averages engineering and research and requires both answer orders to be correct. Unparseable responses count as wrong. Reasoning settings and output budget are part of the setup.

Note (DeepSeek-V4-Flash): 2 of 1,004 responses were unparseable and counted as wrong. Not a Gradually test.

1.GPT-5.6 Sol59.7%