DeepSeek V4 Pro 0813

DeepSeek V4 Pro 0813 is a reasoning model from DeepSeek in the DeepSeek V4 Pro family. 32 benchmarks count toward its score, in 6 categories.

availableShows if the model has enough results for an index.
IndexOverall score out of 100.68.8 ±4.4
CoverageShare of the index weight with results.85%
SpeedOutput tokens per second.68/s
Input / 1MUS dollars per 1M input tokens.$0.435
Output / 1MUS dollars per 1M output tokens.$0.87
ContextMaximum tokens in one request.1M
EloLMArena rating and rank.N/A

The index is a score out of 100. The ± range shows how much it can change.

CapabilitiesScore per category, out of 100.

Out of 100
AgenticMulti-step tasks with tools.
70.3
CodingCode writing and repair.
65.4
ReasoningLogic problems and puzzles.
70.0
MultimodalTasks with images and text.
N/A
KnowledgeFacts and expert knowledge.
68.7
MultilingualTasks in many languages.
N/A
InstructionTasks with strict rules in the prompt.
66.7
MathMath problems.
61.4

Results

32 counted
BenchmarkThe test name.CategoryThe capability that the test measures.ResultThe score from the publisher.IndexThis result as a score out of 100.RunThe settings of the run.DateDate of the result.Published byThe source of the result.
τ²-Bench Tool-Agent-User EvaluationAgentic96.2%67.8Victor Barres et al.
Harvard-MIT Mathematics Tournament February 2026Math95.2%61.4Qwen
LiveCodeBench Pass@1 with Chain-of-ThoughtCoding93.5%DeepSeek
Artificial Analysis GPQA DiamondKnowledge92.8%63.7Artificial Analysis
Apex ShortlistMath90.2%DeepSeek-AI
Graduate-Level Google-Proof Q&AKnowledge90.1%61.4David Rein et al.
GPQA DiamondKnowledge90.1%61.4David Rein et al.
IMOAnswerBenchMath89.8%DeepSeek-AI
Terminal-Bench 2.1 (provider run)Agentic87.9%75.9DeepSeek-AI
Terminal-Bench 2.1 (provider run)Agentic87.9%75.9DeepSeek-AI
Massive Multitask Language Understanding ProfessionalKnowledge87.5%58.4Yubo Wang et al.
Chinese-SimpleQAKnowledge84.4%DeepSeek-AI
MRCR 1MReasoning83.5%DeepSeek-AI
BrowseCompAgentic83.4%69.8OpenAI
CyberGymAgentic83.3%73.2Zhun Wang et al.
Software Engineering Benchmark VerifiedCoding80.6%62.2Carlos E. Jimenez et al.
Artificial Analysis Long Context ReasoningReasoning80.3%63.8Artificial Analysis
Artificial Analysis IFBenchInstruction76.5%66.7Artificial Analysis
Toolathlon-VerifiedAgentic74.1%72.3Moonshot AI
MCP AtlasAgentic73.6%63.6OpenAI
DeepSeek DSBench FullStackCoding71.1%DeepSeek-AI
Artificial Analysis Coding IndexCoding68.8%67.5Artificial Analysis
DeepSeek DSBench HardCoding67.2%DeepSeek-AI
DeepSWEAgentic62.7%69.4Datacurve AI
CorpusQA 1MReasoning62.0%DeepSeek-AI
NL2RepoCoding61.5%73.4MiniMax
Humanity's Last Exam with toolsAgentic60.0%72.7DeepSeek-AI
OpenHarmony Bench v1.0Coding59.0%68.1OpenHarmony Bench authors
Measuring Short-Form Factuality in Large Language ModelsKnowledge57.9%Jason Wei et al.
SWE-bench ProCoding55.4%57.6Xiang Deng et al.
GDPval-AA normalizedAgentic54.5%77.4Artificial Analysis
Artificial Analysis Intelligence IndexKnowledge53.2%88.9Artificial Analysis
ToolathlonAgentic51.8%66.3OpenAI
Artificial Analysis SciCodeCoding51.0%63.6Artificial Analysis
Artificial Analysis EnterpriseOps-GymAgentic49.6%75.2Artificial Analysis
Artificial Analysis Agentic IndexAgentic49.6%77.6Artificial Analysis
Artificial Analysis Omniscience AccuracyKnowledge49.1%74.6Artificial Analysis
Humanity's Last ExamKnowledge42.7%65.0Center for AI Safety et al.
Artificial Analysis Humanity's Last ExamKnowledge41.0%69.0Artificial Analysis
ApexMath38.3%DeepSeek-AI
AutomationBenchAgentic31.8%68.4Moonshot AI
Agents' Last ExamAgentic25.7%65.4DeepSeek-AI
APEX-Agents-AAAgentic24.3%60.3Artificial Analysis / Mercor
Critical Physics TasksReasoning18.0%76.1Artificial Analysis

32 benchmarks count, from 34 of 44 results. A grey row does not count. Too few models took that benchmark.

Sources

BenchLM benchmark aggregationUsed with attribution; per-benchmark results credited to their original authors

Same level, lower price

Qwen3.8-Flash-Next66.4 · FreeGPT-6 Luna68.4 · $0.5DeepSeek V4.1 Flash70.6 · $0.6

More from DeepSeek

DeepSeek V4.1 Flash70.6DeepSeek V4 Flash 073164.6DeepSeek V4 Pro 042362.4DeepSeek V4 Flash 042363.0DeepSeek V3.2 (Thinking)49.9