RankingEffort benchmarks

DeepSeek V4-Pro 0424 effort benchmarks

Compare measured Low, Medium, High and other settings. Every number keeps its source.

Model

DeepSeek V4-Pro 0424

48 measured results · 3 reported effort labels · 27 benchmark/harness combinations. Model evidence and sources →

Capability score evidence

DeepSeek V4-Pro 0424 · Max: 34.9 overall comparison score. Meets the overall evidence requirements.

Which sources carry the weight?

Artificial Analysis · 100.0% direct comparison weight · independent

These shares describe the comparisons entering the fit, not fractions of the final score.

How sensitive is this score?

Remove one evaluator or family and refit the graph against the same overall panel. Values use the matching dataset audited 2026-09-12. These are sensitivity checks, not confidence intervals; ranking eligibility may also change.

Removed evidence	Recomputed score	Change
Artificial Analysis	Comparison path lost	—
ARC Prize	38.4	+3.5
Datacurve	31.4	-3.4
LiveCodeBench	34.7	-0.1
Scale	36.4	+1.5
Terminal-Bench	34.9	+0.1
Cognition	35.5	+0.6
LiveBench	35.1	+0.3
SWE-rebench	35.3	+0.4
Epoch AI	34.8	-0.1
Vals AI	34.0	-0.8
Provider reports · Anthropic	34.9	+0.0
Provider reports · Moonshot AI	34.8	-0.0
Provider reports · Meta	34.8	-0.1
Provider reports · Mistral	34.9	+0.0
All provider reports	34.8	-0.1

Every benchmark family

Removed evidence	Recomputed score	Change
arc-agi	38.3	+3.5
deepswe	31.4	-3.4
automationbench	33.7	-1.2
mmmu-pro	35.8	+1.0
epoch-game-puzzles	34.0	-0.8
briefcase	35.6	+0.7
apex-agents	35.6	+0.7
simpleqa	35.6	+0.7
frontiercode	35.5	+0.6
analyst-agent	35.5	+0.6
ifbench	34.3	-0.6
hle	35.4	+0.6
omniscience	35.4	+0.5
harvey	34.4	-0.4
swe-rebench	35.3	+0.4
mmlu-pro	35.3	+0.4
aa-lcr	34.5	-0.4
gpqa	34.5	-0.4
gdp-pdf	35.2	+0.4
enterprise-ops	35.2	+0.3
itbench	34.6	-0.3
multi-swe-bench	34.6	-0.3
tau-bench	34.6	-0.3
livebench-instructions	35.1	+0.2
livebench-language	35.1	+0.2
scicode	35.1	+0.2
swe-atlas-qna	34.7	-0.2
frontiermath	35.0	+0.2
arc-agi-3	34.7	-0.2
gdpval	35.0	+0.2
livecodebench	34.7	-0.1
vibe-code	34.7	-0.1
livebench-coding	34.7	-0.1
terminal-bench	34.7	-0.1
livebench-reasoning	34.8	-0.1
swe-atlas-test-writing	34.8	-0.1
critpt	35.0	+0.1
vals-finance-agent	34.8	-0.1
vals-legal-research	34.8	-0.1
vals-code-migration	34.8	-0.1
gmmlu	34.9	+0.0
swe-atlas-refactoring	34.8	-0.0
livebench-data	34.9	+0.0
livebench-math	34.9	+0.0
aime	34.8	-0.0
vals-excel-modeling	34.9	-0.0
math500	34.9	-0.0
swe-bench-pro	34.9	+0.0
enigma-eval	34.9	+0.0

What evidence is missing from the fit?

Across all collected settings for this model: 48 observations, 45 contributing, 0 matched without graph weight and 3 excluded. Counts do not establish rank eligibility.

3 · Benchmark family or source protocol has not been reviewed

Inspect every observation and exclusion →

Capability profile

One configuration, seven capabilities

Comparison score · 0–100

ADeepSeek V4-Pro 0424 · Max4 of 7 capabilities supported

Gaps are unknown, not zero. Hollow points are preliminary.

Agentic
A38.4Supported
Hard reasoning
A37.7Supported
Coding
A40.2Supported
Human pref
Unknown
Knowledge
A39.7Supported
Multimodal
Unknown
Long context
A51.7Preliminary

Each profile belongs to the named configuration. Capability scores use a shared panel for all axes, distinct from the overall ranking panel; they are not accuracy percentages or direct head-to-head wins.

Compare this configuration →Explore the model evidence profile →

These are published benchmark scores. Supported effort estimates also appear on the Capability leaderboard. Results from different harnesses occupy different rows. Missing cells stay unknown; no scores are copied between effort levels.

Benchmark / source	High	Max	Non-reasoning
AA-AnalystAgentArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Share of tasks solved on all 5 of 5 attempts · 80 tasks, 5 attempts per task · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	18.75% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	Not reported
AA-Briefcase EloArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details AA-Briefcase is an agentic knowledge work benchmark developed by Artificial Analysis. AA-Briefcase Elo is a combined metric that aggregates rubric pass rate, analytical quality Elo and presentation Elo · Higher is better · Evaluation results measured independently by Artificial Analysis Unit: elo. Source collected 2026-09-12.	Not reported	924.56 Elo Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	Not reported
APEX-Agents-AAArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	24.26% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	Not reported
AA-LCR v1.1Artificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	70.33% Reported settings DeepSeek V4 Pro (Reasoning, High Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-high	74.67% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	53% Reported settings DeepSeek V4 Pro (Non-reasoning); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-non-reasoning
AutomationBench-AAArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Share of task objectives completed with no guardrail violations · Higher is better · Benchmark developed by Zapier · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	56.29% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	Not reported
CritPtArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Benchmark developed by Argonne and UIUC, with contributions from 60+ researchers globally · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	10% Reported settings DeepSeek V4 Pro (Reasoning, High Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-high	12.86% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	0.86% Reported settings DeepSeek V4 Pro (Non-reasoning); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-non-reasoning
EnterpriseOps-Gym-AAArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Overall task success rate across all eight enterprise domains (oracle tool mode) · Benchmark developed by ServiceNow Research · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	40.44% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	Not reported
GDP.pdf: All-passArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Share of attempts where every atomic criterion passed · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	13.4% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	Not reported
GDPval-AA v2 LeaderboardArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Elo rating for performance on real-world work tasks · Anchored to a human baseline of 1,000 · Higher is better · Evaluation results measured independently by Artificial Analysis Unit: elo. Source collected 2026-09-12.	1,218.72 Elo Reported settings DeepSeek V4 Pro (Reasoning, High Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-high	1,223.37 Elo Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	Not reported
GPQA DiamondArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	90.51% Reported settings DeepSeek V4 Pro (Reasoning, High Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-high	88.79% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	71.72% Reported settings DeepSeek V4 Pro (Non-reasoning); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-non-reasoning
Harvey LAB-AA: Criterion Pass RateArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Share of atomic pass/fail rubric criteria the deliverables satisfy, graded by a single LLM judge · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	84.4% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	Not reported
Humanity's Last ExamArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	35.22% Reported settings DeepSeek V4 Pro (Reasoning, High Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-high	37.53% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	8.25% Reported settings DeepSeek V4 Pro (Non-reasoning); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-non-reasoning
IFBenchArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	71.29% Reported settings DeepSeek V4 Pro (Reasoning, High Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-high	76.46% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	45.78% Reported settings DeepSeek V4 Pro (Non-reasoning); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-non-reasoning
ITBench-AAArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Combined public and private Site Reliability Engineering (SRE) tasks · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	38.32% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	Not reported
MLCR-AAArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Share of tasks judged accurate, complete, and concise · Expert + compound (hard) sets · Benchmark developed by Wisedocs · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	21.11% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	Not reported
AA-Omniscience IndexArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details AA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct. · Evaluation results measured independently by Artificial Analysis Unit: index. Source collected 2026-09-12.	-10.57 Reported settings DeepSeek V4 Pro (Reasoning, High Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-high	-10.73 Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	-29.87 Reported settings DeepSeek V4 Pro (Non-reasoning); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-non-reasoning
AA-Omniscience AccuracyArtificial Analysis · 6000 questions; 1 attempt; correct-answer share; GPT-5.6 Luna medium grader Benchmark details AA-Omniscience Accuracy (higher is better) measures the proportion of correctly answered questions out of all questions, regardless of whether the model chooses to answer · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	41.4% Reported settings DeepSeek V4 Pro (Reasoning, High Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-high	42.95% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	30.87% Reported settings DeepSeek V4 Pro (Non-reasoning); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-non-reasoning
SciCodeArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	50.81% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	Not reported
𝜏²-Bench TelecomArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	94.15% Reported settings DeepSeek V4 Pro (Reasoning, High Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-high	96.2% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	91.23% Reported settings DeepSeek V4 Pro (Non-reasoning); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-non-reasoning
𝜏³-BankingArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Benchmark developed by Sierra Research · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	26.19% Reported settings DeepSeek V4 Pro (Reasoning, High Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-high	30.1% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	Not reported
Terminal-Bench HardArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	41.67% Reported settings DeepSeek V4 Pro (Reasoning, High Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-high	46.21% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	36.36% Reported settings DeepSeek V4 Pro (Non-reasoning); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-non-reasoning
Terminal-Bench v2.1Artificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Benchmark developed by the Laude Institute and open-source contributors · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	64.79% Reported settings DeepSeek V4 Pro (Reasoning, High Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424-high	64.04% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	Not reported
Terminal-Bench v4.0Artificial Analysis · mini-SWE-agent v2.4.6; 66 tasks; pass@1 averaged over 3 repeats; 500 steps; task timeout up to 8 hours Benchmark details Benchmark developed by the Laude Institute, Stanford, and open-source contributors · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	14.65% Reported settings DeepSeek V4 Pro (Reasoning, Max Effort); Model source: https://artificialanalysis.ai/models/deepseek-v4-pro-0424	Not reported
SWE-rebench · 2026-03 · toolsSWE-rebench · SWE-rebench standardized ReAct tools Benchmark details Average resolved rate across repeated runs, not best-of-N success. Disjoint complete calendar months within each model’s tested range; standardized harness and all-language task set. Unit: percent. Source collected 2026-09-12.	46.72% Reported settings SWE-rebench standardized ReAct tools; swe-rebench-2026-03-tools; DeepSeek-V4 Pro [high]; checkpoint=deepseek-v4-pro-0424	Not reported	Not reported
SWE-rebench · 2026-04 · toolsSWE-rebench · SWE-rebench standardized ReAct tools Benchmark details Average resolved rate across repeated runs, not best-of-N success. Disjoint complete calendar months within each model’s tested range; standardized harness and all-language task set. Unit: percent. Source collected 2026-09-12.	41.58% Reported settings SWE-rebench standardized ReAct tools; swe-rebench-2026-04-tools; DeepSeek-V4 Pro [high]; checkpoint=deepseek-v4-pro-0424	Not reported	Not reported
SWE-rebench · 2026-05 · toolsSWE-rebench · SWE-rebench standardized ReAct tools Benchmark details Average resolved rate across repeated runs, not best-of-N success. Disjoint complete calendar months within each model’s tested range; standardized harness and all-language task set. Unit: percent. Source collected 2026-09-12.	41.56% Reported settings SWE-rebench standardized ReAct tools; swe-rebench-2026-05-tools; DeepSeek-V4 Pro [high]; checkpoint=deepseek-v4-pro-0424	Not reported	Not reported
SWE-rebench · 2026-06 · toolsSWE-rebench · SWE-rebench standardized ReAct tools Benchmark details Average resolved rate across repeated runs, not best-of-N success. Disjoint complete calendar months within each model’s tested range; standardized harness and all-language task set. Unit: percent. Source collected 2026-09-12.	35.24% Reported settings SWE-rebench standardized ReAct tools; swe-rebench-2026-06-tools; DeepSeek-V4 Pro [high]; checkpoint=deepseek-v4-pro-0424	Not reported	Not reported

What these comparisons mean

Each source reports its own effort labels. Matching labels do not establish equal compute budgets or identical fallback behavior. Open Reported settings to inspect the exact model name, agent and task coverage. Multiple results at one level remain visible rather than selecting the best. Source-reported None or Non-reasoning labels are not treated as evidence that an API supports those settings.

These measurements use their own dated source collection. The leaderboard fits identified configurations jointly; it contains no pooled model ratings. Explicit thinking budgets are separate configurations. Read the methodology.

Download all effort benchmark results · 7,424 results across 240 models · Collected 2026-09-12

DeepSeek V4-Pro 0424 effort benchmarks