Skip to main content

Model profile · DeepSeek

DeepSeek V4 Flash (Max)

CurrentReleased Apr 24, 2026Open WeightReasoning1M context
DeepSeek V4 Flash (Max) scores 51.8 out of 100 and ranks #108 of 215. This profile shows 23 source-displayable benchmark rows; its strongest eligible category is Knowledge at #42. API pricing is $0.14 input and $0.28 output per million tokens, with cached input at $0.0028.

Data as of July 28, 2026 · How the score is built

Strongest published evidence

Knowledge ranks #42. Particularly effective for knowledge-intensive tasks like research, analysis, and factual Q&A.

Validate before choosing

23 published rows leave some tracked benchmark slots empty. Independent runtime speed has not been measured.

Decision snapshot

Each value carries a field reference instead of floating alone. Markers compare this model with the current ranked and priced catalog; they are not absolute quality thresholds.

Capability

51.8/100

field median 57.2

#108 of 215 ranked models

Price

$0.14input / $0.28 output

input median $1

cached $0.003 · blended $0.21

Speed

Not measured

field median 107 tok/s

Time to first token not measured

Context

1Mtokens

field median 256,000

Maximum output length is tracked separately

Capability shape

Each axis shows percentile within that category’s eligible cohort. The comparison outline is the median of the six nearest public-score peers; a collapsed vertex means the category is not rank-eligible.

DeepSeek V4 Flash (Max) category percentile values

  • Agentic61st percentile
  • Coding59th percentile
  • ReasoningNot eligible
  • Knowledge24th percentile
  • MathNot eligible
  • MultilingualNot eligible
  • MultimodalNot eligible
  • Instruction followingNot eligible

The dashed outline is median of 6 nearest peers.

Eligible category ranks

  1. Agentic#51/129
  2. Coding#54/130
  3. ReasoningNot ranked
  4. Knowledge#42/55
  5. MathNot ranked
  6. MultilingualNot ranked
  7. MultimodalNot ranked
  8. Inst. FollowingNot ranked
Top decileTop quartileMid-fieldNot eligible

What it costs to get this score

Published API price against the public score. The x-axis uses a log scale; the dashed path marks models that are not beaten by a cheaper, higher-scoring option. Price uses average of published input and output rates.

Explore all models

The chart opens on the current model. Scroll horizontally to inspect the full price axis.

Current modelDeepSeek V4 Flash (Max) · 51.8 score · $0.21 blended per million tokens
405060708090$0.50$1$5$10$25↘ frontierDeepSeek V4 Flash (Max)

Horizontal: blended price per million tokens, log scale · Vertical: public score

How much of this is verified

Coverage is split by category so a strong number never hides a thin evidence base. Verified means the row is tied to a published source; provisional rows remain visible but separate.

  1. Agentic5/5 verified
  2. Coding6/6 verified
  3. Reasoning2/2 verified
  4. Knowledge6/6 verified
  5. Math4/4 verified
  6. MultilingualNot measured
  7. MultimodalNot measured
  8. Inst. FollowingNot measured
Verified sourceProvisionalNot measured

Spec sheet

Each documented value carries its source. Missing fields stay visible as not sourced or not published, rather than disappearing from the page.

API model ID
deepseek-v4-flashDeepSeek models and pricing
Maximum output
Not sourced yet
Knowledge cutoff
Not sourced yet
Parameters
Not sourced yet
Availability
DeepSeek API · open weightsDeepSeek V4 release
Cloud regions
Not tracked yet
Lifecycle
activeDeepSeek V4 release
API capabilities
Tool calling, structured outputs, and batch support are not tracked yet
Prompt caching
Published at $0.003 per million cached input tokens
Self-host
Open weights available; hardware estimate not sourced
Rate limits
Not tracked yet

Deployment options

Self-host and provider-specific paths stay separate from benchmark evidence so operating constraints are visible before a score becomes the whole decision.

Published weights are available, but BenchLM does not yet have a sourced parameter and VRAM profile for this exact model. Hardware cost estimates stay unavailable until that sizing record is complete.

Estimate VRAM from known parameters

Category score record

Scores and ranks appear only where published evidence can be displayed. The table keeps the score, weight, cohort, and evidence state together.

Category scores, ranks, weighting, benchmark coverage, and evidence status
CategoryScore
AgenticRank #51 of 129Percentile 61stWeight 22%5 benchmarksVerified49.0
CodingRank #54 of 130Percentile 59thWeight 20%6 benchmarksVerified51.4
ReasoningWeight 17%2 benchmarksVerifiedScore pending
KnowledgeRank #42 of 55Percentile 24thWeight 12%6 benchmarksVerified60.1
MathRank Not rankedWeight 5%4 benchmarksVerified81.6
MultilingualWeight 7%0 benchmarksNot measuredNot measured
MultimodalWeight 12%0 benchmarksNot measuredNot measured
Inst. FollowingWeight 5%0 benchmarksNot measuredNot measured

Benchmark ledger

Coding opens by default. The marker compares each value with the best source-verified result in the catalog; provisional leaders do not set the reference. Expand the remaining categories for every published row.

Coding6 rows
Coding benchmark values, best verified comparison, weight, and source status
SWE-bench VerifiedSoftware Engineering Benchmark VerifiedScore79%Versus best verified row

Best verified: Claude Opus 5 · 96%

Gap17 behindWeightWeighted 16%
SWE-bench ProScore52.6%Versus best verified row

Best verified: Claude Mythos 5 · 80.3%

Gap27.7 behindWeightWeighted 10%
LiveCodeBench Pass@1-COTLiveCodeBench Pass@1 with Chain-of-ThoughtScore91.6%Versus best verified row

Best verified: DeepSeek V4 Pro (Max) · 93.5%

Gap1.9 behindWeightDisplay only
CodeforcesCodeforces RatingScore3052.0Versus best verified row

Best verified: DeepSeek V4 Pro (Max) · 3206.0

Gap154 behindWeightDisplay only
SWE MultilingualScore73.3%Versus best verified row

Best verified: Claude Opus 5 · 89.5%

Gap16.2 behindWeightDisplay only
Terminal-Bench 2.0Score56.9%Versus best verified row

Best verified: GPT-5.6 Sol · 91.9%

Gap35 behindWeightDisplay only
Agentic5 rows
Agentic benchmark values, best verified comparison, weight, and source status
Terminal-Bench 2.0Score56.9%Versus best verified row

Best verified: GPT-5.6 Sol · 91.9%

Gap35 behindWeightWeighted 38%
BrowseCompScore73.2%Versus best verified row

Best verified: GPT-5.6 Sol · 92.2%

Gap19 behindWeightWeighted 28%
HLE w/ toolsHumanity's Last Exam with toolsScore45.1%Versus best verified row

Best verified: Claude Opus 5 · 64.7%

Gap19.6 behindWeightDisplay only
MCP AtlasScore69%Versus best verified row

Best verified: Muse Spark 1.1 · 88.1%

Gap19.1 behindWeightDisplay only
ToolathlonScore47.8%Versus best verified row

Best verified: Muse Spark 1.1 · 75.6%

Gap27.8 behindWeightDisplay only
Reasoning2 rows
Reasoning benchmark values, best verified comparison, weight, and source status
MRCR 1MScore78.7%Versus best verified row

Best verified: DeepSeek V4 Pro (Max) · 83.5%

Gap4.8 behindWeightDisplay only
CorpusQA 1MScore60.5%Versus best verified row

Best verified: DeepSeek V4 Pro (Max) · 62.0%

Gap1.5 behindWeightDisplay only
Knowledge6 rows
Knowledge benchmark values, best verified comparison, weight, and source status
HLEHumanity's Last ExamScore34.8%Versus best verified row

Best verified: Claude Opus 5 · 64.7%

Gap29.9 behindWeightWeighted 45%
MMLU-ProMassive Multitask Language Understanding ProfessionalScore86.2%Versus best verified row

Best verified: Qwen3.7 Max · 89.6%

Gap3.4 behindWeightWeighted 30%
SimpleQAMeasuring Short-Form Factuality in Large Language ModelsScore34.1%Versus best verified row

Best verified: DeepSeek V4 Pro (Max) · 57.9%

Gap23.8 behindWeightWeighted 11%
GPQAGraduate-Level Google-Proof Q&AScore88.1%Versus best verified row

Best verified: Sakana Fugu-Ultra · 95.5%

Gap7.4 behindWeightWeighted 7%
Chinese-SimpleQAScore78.9%Versus best verified row

Best verified: DeepSeek V4 Pro (Max) · 84.4%

Gap5.5 behindWeightDisplay only
GPQA-DGPQA DiamondScore88.1%Versus best verified row

Best verified: Sakana Fugu-Ultra · 95.5%

Gap7.4 behindWeightDisplay only
Math4 rows
Math benchmark values, best verified comparison, weight, and source status
HMMT Feb 2026Harvard-MIT Mathematics Tournament February 2026Score94.8%Versus best verified row

Best verified: Qwen3.7 Max · 97.1%

Gap2.3 behindWeightWeighted 25%
IMOAnswerBenchScore88.4%Versus best verified row

Best verified: Qwen3.7 Max · 90.0%

Gap1.6 behindWeightDisplay only
ApexScore33.0%Versus best verified row

Best verified: Qwen3.7 Max · 44.5%

Gap11.5 behindWeightDisplay only
Apex ShortlistScore85.7%Versus best verified row

Best verified: DeepSeek V4 Pro (Max) · 90.2%

Gap4.5 behindWeightDisplay only

Lineage

The sequence follows explicit supersedes links. Scores and prices remain blank when the corresponding public row or first-party rate is unavailable.

flash-reasoning · max

How to read this profile

The visual layer above carries the decisions. These notes preserve the model, ranking, coverage, and family context behind the numbers.

DeepSeek V4 Flash (Max) ranks #108 of 215 on the public leaderboard with a score of 51.76/100. It does not yet have enough sourced coverage for a verified position.

DeepSeek V4 Flash (Max) is a open weight model with a 1M context window. It uses an explicit reasoning mode, which can improve complex problem solving while adding latency and token use.

Official exact-value post-training snapshot from Table 7 of the DeepSeek-V4 technical report. This row maps the Max reasoning-effort column for DeepSeek V4 Flash.

DeepSeek V4 Flash (Max) sits in the DeepSeek V4 family with DeepSeek V4 Pro (Max), DeepSeek V4 Pro (High), DeepSeek V4 Pro, DeepSeek V4 Flash (High), DeepSeek V4 Pro Base, DeepSeek V4 Flash Base, DeepSeek V4 Flash. 23 of 369 tracked benchmark slots currently have displayable evidence. Missing categories stay blank.

Its strongest eligible category is Knowledge at #42, while its lowest eligible position is Coding at #54. particularly effective for knowledge-intensive tasks like research, analysis, and factual Q&A.

Frequently asked questions

How does DeepSeek V4 Flash (Max) perform overall in AI benchmarks?

DeepSeek V4 Flash (Max) ranks #108 out of 215 models on the public BenchAlign leaderboard, with a score of 51.76/100. Its evidence status is Estimated, and this profile shows 23 source-displayable benchmark rows. The label describes evidence depth, not a provider quality claim; inspect category rows before choosing a workload.

Is DeepSeek V4 Flash (Max) good for knowledge and understanding?

DeepSeek V4 Flash (Max) ranks #42 out of 55 eligible models for knowledge and understanding, with a public category score of 60.1/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.

Is DeepSeek V4 Flash (Max) good for coding and programming?

DeepSeek V4 Flash (Max) ranks #54 out of 130 eligible models for coding and programming, with a public category score of 51.4/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.

Is DeepSeek V4 Flash (Max) good for mathematics?

DeepSeek V4 Flash (Max) has source-displayable benchmark coverage for mathematics, but the public category table does not assign it a rank there. The individual rows remain available for inspection. A missing category position means the evidence threshold was not met; it does not convert the model's unmeasured work into a zero.

Is DeepSeek V4 Flash (Max) good for reasoning and logic?

DeepSeek V4 Flash (Max) has source-displayable benchmark coverage for reasoning and logic, but the public category table does not assign it a rank there. The individual rows remain available for inspection. A missing category position means the evidence threshold was not met; it does not convert the model's unmeasured work into a zero.

Is DeepSeek V4 Flash (Max) good for agentic tool use and computer tasks?

DeepSeek V4 Flash (Max) ranks #51 out of 129 eligible models for agentic tool use and computer tasks, with a public category score of 49/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.

Is DeepSeek V4 Flash (Max) open source?

DeepSeek V4 Flash (Max) is an open-weight model from DeepSeek. Its weights can be downloaded for local or hosted deployment, subject to the published license. Open weight does not automatically mean open source: training data and training code may remain private, and commercial restrictions can still apply.

Which sibling models are related to DeepSeek V4 Flash (Max)?

DeepSeek V4 Flash (Max) belongs to the DeepSeek V4 family. Related tracked variants include DeepSeek V4 Pro (Max), DeepSeek V4 Pro (High), DeepSeek V4 Pro, plus 4 more. A sibling link indicates shared lineage or a documented configuration relationship; it does not mean the variants have identical pricing, context limits, benchmark evidence, or deployment behavior. Compare before switching.

Does DeepSeek V4 Flash (Max) have full benchmark coverage on BenchLM?

No. DeepSeek V4 Flash (Max) currently has 47 source-displayable rows across 369 tracked benchmark slots. The profile exposes published, non-generated evidence and leaves missing categories blank until an exact evaluation is available. Coverage describes how much was measured; it is not a penalty added to an individual benchmark result.

What is the context window size of DeepSeek V4 Flash (Max)?

DeepSeek V4 Flash (Max) has a documented context window of 1M. That figure is the maximum combined prompt and retained-conversation space reported for this exact model; it is not the maximum output length. The profile keeps output limits separate because providers often publish those limits independently.

Last updated July 28, 2026. Runtime fields remain blank until a sourced snapshot exists.

Make the right model choice this week

The model to choose, the cheaper alternative, and the release we would wait on.