New: Agent Index for knowledge work is live.Explore the index ↗
Compare models
DeepSeek (opens in a new tab)

DeepSeek V4 Pro 0424

Across the task benchmarks we track, DeepSeek V4 Pro 0424’s strongest relative results are in coding, web research, and work automations. On SWE-bench Verified, it scores 80.6%, ranking #1 among 7 models with reported results in our dataset.

Updated: Data as of:
Compare this model
🥇

Strongest results

Strongest results for DeepSeek V4 Pro 0424 in tracked evaluation cohorts
BenchmarkMeasuresScore
SWE-bench ↗Verified / published open-model reportsCoding80.6%#1 of 7
IFBench ↗58 constraint typesPrecise instruction following76.46%#6 of 17
BrowseComp ↗Published open-model reportsWeb research83.4%#4 of 8
AutomationBench-AA ↗1.0.6 / AA private 657Work automations56.29%#25 of 55

Model details

Open source / open weights13 benchmarks
Specifications for DeepSeek V4 Pro 0424
ModelDeepSeek V4 Pro 0424
Agent Index v1.2Insufficient coverage (3/7) · Preview
CreatorDeepSeek (opens in a new tab)
AvailabilityOpen source / open weights
Context size1,000,000 tokens (opens in a new tab)
Max outputNot reported
Knowledge cutoffNot reported
LicenseMIT · Model weights ↗ (opens in a new tab)

Pricing and performance

Per provider and reported configuration. Prices are USD per 1M tokens.

No provider pricing or performance reports have been added for this model yet.

Benchmark performance

DeepSeek V4 Pro 0424 in color, alongside leading models. Latest stored reports, best effort per report date.

Reported benchmarks

All published results in our dataset, with versions, evaluation settings, and sources.

All published benchmark reports for DeepSeek V4 Pro 0424
BenchmarkScoreEvaluation setupSource
AA-Briefcase1.1 / combined EloCombined Elo · Higher is better921.93 Eloselection: best reported configuration per release · DeepSeek V4 Pro 0424 (Max) · tools: benchmark-definedArtificial Analysis (opens in a new tab)independent
AutomationBench-AA1.0.6 / AA private 657Objectives completed without guardrail violations · Higher is better56.29%DeepSeek V4 Pro 0424 (Max); Private 657 tasks; 50 turns; objective completionArtificial Analysis (opens in a new tab)independent
BrowseCompPublished open-model reportsAccuracy · Higher is better83.4%harness comparison: provider-reported; settings differ across models · Think Max; provider browsing agent; discard-all context management; original V4 releaseDeepSeek ↗ (opens in a new tab)provider reported
GDP.pdf100 tasks / All-pass / five attemptsAccuracy · Higher is better13.4%selection: best reported configuration per release · DeepSeek V4 Pro 0424 (Max) · tools: benchmark-definedArtificial Analysis (opens in a new tab)independent
GDPval-AA2.1Elo · Higher is better1,166.24 EloDeepSeek V4 Pro 0424 (Max); 220 tasks; Stirrup agent; pairwise EloArtificial Analysis (opens in a new tab)independent
GDPval-AA2.1Elo · Higher is better1,159.95 EloDeepSeek V4 Pro 0424 (High); 220 tasks; Stirrup agent; pairwise EloArtificial Analysis (opens in a new tab)independent
GPQADiamond / Artificial AnalysisAccuracy · Higher is better88.788%DeepSeek V4 Pro 0424 (Max); 198-question Diamond subset; independent AA evaluationArtificial Analysis (opens in a new tab)independent
GPQADiamond / Artificial AnalysisAccuracy · Higher is better71.717%DeepSeek V4 Pro 0424 (Non-reasoning); 198-question Diamond subset; independent AA evaluationArtificial Analysis (opens in a new tab)independent
GPQADiamond / Artificial AnalysisAccuracy · Higher is better90.505%DeepSeek V4 Pro 0424 (High); 198-question Diamond subset; independent AA evaluationArtificial Analysis (opens in a new tab)independent
Humanity's Last ExamHLE — source does not specify subsetAccuracy · Higher is better48.2%Original V4 Think Max; HLE with tools; source does not explicitly identify subset · Tools enabledDeepSeek ↗ (opens in a new tab)provider reported
Humanity's Last ExamStable HLE — text-only subsetAccuracy · Higher is better37.535%DeepSeek V4 Pro 0424 (Max); original HLE text-only 2,158 questions; no tools; not HLE Diamond · No toolsArtificial Analysis (opens in a new tab)independent
Humanity's Last ExamStable HLE — text-only subsetAccuracy · Higher is better8.248%DeepSeek V4 Pro 0424 (Non-reasoning); original HLE text-only 2,158 questions; no tools; not HLE Diamond · No toolsArtificial Analysis (opens in a new tab)independent
Humanity's Last ExamStable HLE — text-only subsetAccuracy · Higher is better35.218%DeepSeek V4 Pro 0424 (High); original HLE text-only 2,158 questions; no tools; not HLE Diamond · No toolsArtificial Analysis (opens in a new tab)independent
IFBench58 constraint typesAccuracy · Higher is better76.463%selection: best reported configuration per release · DeepSeek V4 Pro 0424 (Max) · tools: benchmark-definedArtificial Analysis (opens in a new tab)independent
SWE-benchVerified / published open-model reportsResolved · Higher is better80.6%harness comparison: provider-reported; settings differ across models · DeepSeek original V4 release; Think Max; published SWE Verified resolved score; original release, not 0813DeepSeek ↗ (opens in a new tab)provider reported
Terminal-Bench2.0 / published open-model reportsTask success · Higher is better67.9%harness comparison: provider-reported; settings differ across models · Published Terminal-Bench 2.0 score; not 2.1DeepSeek ↗ (opens in a new tab)provider reported
Terminal-Bench2.1 / Artificial AnalysisPass@1 · Higher is better64.045%DeepSeek V4 Pro 0424 (Max); 89 tasks; Terminus 2; e2b; three repeatsArtificial Analysis (opens in a new tab)independent
Terminal-Bench2.1 / Artificial AnalysisPass@1 · Higher is better64.794%DeepSeek V4 Pro 0424 (High); 89 tasks; Terminus 2; e2b; three repeatsArtificial Analysis (opens in a new tab)independent
Terminal-Bench4.0 / mini-swe-agentTask success · Higher is better14.646%harness: mini-swe-agent · repeats: 3 · selection: best reported configuration per release · DeepSeek V4 Pro 0424 (Max)Artificial Analysis (opens in a new tab)independent