DeepSeek model product
DeepSeek-V4-Pro
DeepSeek-V4-Pro-0813 is DeepSeek’s GA production release of DeepSeek-V4-Pro for agentic coding and tool-use workloads.
Updated Sep 8, 2026. Default version: DeepSeek-V4-Pro-0813
LLMBoard Score86.8DeepSeek-V4-Pro-0813
Coverage20%11 benchmark families
Context window1MTokens
Official input priceN/AOfficial price unavailable
DeepSeek-V4-Pro Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
DeepSeek-V4-Pro Benchmark Results
Benchmark scores for DeepSeek-V4-Pro-0813.
Sort by
Benchmark | Score | Rank | Participants | Percentile | Evidence | Evaluated |
|---|
| BenchmarkDSBench-FullStack | Score71.10% | Rank01 | Participants2 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkDSBench-Hard | Score67.20% | Rank01 | Participants3 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkHumanity's Last Exam (with tools, text-only) | Score60.00% | Rank01 | Participants5 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkNL2Repo | Score61.50% | Rank01 | Participants22 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkCyberGym | Score83.30% | Rank03 | Participants15 | Percentile85.71% | EvidenceC | Evaluated |
| BenchmarkHumanity's Last Exam (no tools, text-only) | Score42.70% | Rank03 | Participants3 | Percentile0.00% | EvidenceC | Evaluated |
| BenchmarkToolathlon | Score74.10% | Rank03 | Participants41 | Percentile95.00% | EvidenceC | Evaluated |
| BenchmarkLM Arena Agent Tool Hallucination | Score0.78% | Rank04 | Participants49 | Percentile93.75% | EvidenceA | Evaluated |
| BenchmarkAutomationBench | Score31.80% | Rank06 | Participants18 | Percentile70.59% | EvidenceC | Evaluated |
| BenchmarkDeepSWE | Score62.70% | Rank06 | Participants13 | Percentile58.33% | EvidenceC | Evaluated |
| BenchmarkTerminal-Bench 2.1 | Score87.90% | Rank06 | Participants35 | Percentile85.29% | EvidenceC | Evaluated |
| BenchmarkHumanity's Last Exam | Score60.00% | Rank07 | Participants103 | Percentile94.12% | EvidenceC | Evaluated |
| BenchmarkLM Arena Agent Task Outcome Explicit | Score11.75% | Rank08 | Participants49 | Percentile85.42% | EvidenceA | Evaluated |
| BenchmarkLM Arena Agent Bash Recovery Steps | Score8.70% | Rank12 | Participants49 | Percentile77.08% | EvidenceA | Evaluated |
| BenchmarkAgents' Last Exam | Score25.70% | Rank14 | Participants16 | Percentile13.33% | EvidenceC | Evaluated |
| BenchmarkLM Arena Agent Leaderboard | Score5.43% | Rank15 | Participants49 | Percentile70.83% | EvidenceA | Evaluated |
| BenchmarkLM Arena Agent Praise Complaint | Score5.24% | Rank17 | Participants49 | Percentile66.67% | EvidenceA | Evaluated |
| BenchmarkLM Arena Webdev | Score1,581.68 rating | Rank17 | Participants97 | Percentile83.33% | EvidenceA | Evaluated |
| BenchmarkLM Arena Agent Steerability | Score0.67% | Rank19 | Participants49 | Percentile62.50% | EvidenceA | Evaluated |
| BenchmarkLM Arena Text | Score1,451.07 rating | Rank33 | Participants210 | Percentile84.69% | EvidenceA | Evaluated |
| BenchmarkLM Arena Text Factuality | Score1,454.90 rating | Rank38 | Participants121 | Percentile69.17% | EvidenceA | Evaluated |
| BenchmarkLM Arena Text Style Control | Score1,459.61 rating | Rank39 | Participants210 | Percentile81.82% | EvidenceA | Evaluated |
DeepSeek-V4-Pro Arena Results
Preference and agent-evaluation results for the default version.
Sort by
Arena | Category | Rank | Rating / score | Votes | Observations | Result date |
|---|
| Arenaagent tool hallucination | Categoryoverall | Rank04 | Rating / score0.01 | VotesN/A | Observations2.7M | Result date |
| Arenatext factuality | Categoryjapanese | Rank06 | Rating / score1,463.81 | Votes489 | ObservationsN/A | Result date |
| Arenaagent task outcome explicit | Categoryoverall | Rank08 | Rating / score0.12 | VotesN/A | Observations20.4K | Result date |
| Arenaagent bash recovery steps | Categoryoverall | Rank12 | Rating / score0.09 | VotesN/A | Observations68K | Result date |
| Arenatext factuality | Categorygerman | Rank12 | Rating / score1,440.08 | Votes517 | ObservationsN/A | Result date |
| Arenaagent | Categoryoverall | Rank15 | Rating / score0.05 | VotesN/A | Observations2.8M | Result date |
| Arenatext style control | Categoryjapanese | Rank15 | Rating / score1,461.87 | Votes696 | ObservationsN/A | Result date |
| Arenatext factuality | Categorykorean | Rank16 | Rating / score1,413.56 | Votes621 | ObservationsN/A | Result date |
| Arenawebdev | Categorywebdev-react | Rank16 | Rating / score1,586.05 | Votes2,319 | ObservationsN/A | Result date |
| Arenaagent praise complaint | Categoryoverall | Rank17 | Rating / score0.05 | VotesN/A | Observations7.2K | Result date |
| Arenawebdev | Categoryoverall | Rank17 | Rating / score1,581.68 | Votes3,518 | ObservationsN/A | Result date |
| Arenawebdev | Categorywebdev | Rank17 | Rating / score1,581.68 | Votes3,518 | ObservationsN/A | Result date |
| Arenatext | Categoryjapanese | Rank18 | Rating / score1,451.16 | Votes696 | ObservationsN/A | Result date |
| Arenaagent steerability | Categoryoverall | Rank19 | Rating / score0.01 | VotesN/A | Observations22K | Result date |
| Arenatext factuality | Categoryindustry medicine and healthcare | Rank19 | Rating / score1,491.18 | Votes3,070 | ObservationsN/A | Result date |
| Arenatext | Categorypolish | Rank20 | Rating / score1,463.60 | Votes1,085 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryspanish | Rank20 | Rating / score1,453.45 | Votes1,390 | ObservationsN/A | Result date |
| Arenawebdev | Categorywebdev-html | Rank20 | Rating / score1,555.52 | Votes388 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryfrench | Rank21 | Rating / score1,485.15 | Votes1,696 | ObservationsN/A | Result date |
| Arenatext factuality | Categorymulti turn | Rank22 | Rating / score1,477.42 | Votes8,554 | ObservationsN/A | Result date |
| Arenatext | Categorymulti turn | Rank25 | Rating / score1,468.23 | Votes9,809 | ObservationsN/A | Result date |
| Arenatext | Categorycreative writing | Rank26 | Rating / score1,444.15 | Votes9,029 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryindustry legal and government | Rank26 | Rating / score1,477.94 | Votes3,547 | ObservationsN/A | Result date |
| Arenatext style control | Categoryindustry business and management and financial operations | Rank26 | Rating / score1,469.18 | Votes999 | ObservationsN/A | Result date |
| Arenatext factuality | Categorycreative writing | Rank27 | Rating / score1,445.17 | Votes8,016 | ObservationsN/A | Result date |
| Arenatext style control | Categorygerman | Rank27 | Rating / score1,468.15 | Votes965 | ObservationsN/A | Result date |
| Arenatext | Categorygerman | Rank28 | Rating / score1,457.02 | Votes965 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryexpert | Rank28 | Rating / score1,489.29 | Votes4,618 | ObservationsN/A | Result date |
| Arenatext | Categoryindustry writing and literature and language | Rank29 | Rating / score1,446.36 | Votes13,158 | ObservationsN/A | Result date |
| Arenatext factuality | Categoryindustry life and physical and social science | Rank29 | Rating / score1,483.96 | Votes7,616 | ObservationsN/A | Result date |
DeepSeek-V4-Pro Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- N/A
- Official provider
- N/A
- Lowest third-party
- From $0.35 input, $0.80 output per 1M via CrofAI
- Tracked offerings
- 30
Sort by
Provider | Provider model ID | Region | Input / 1M | Output / 1M | Context | Updated |
|---|
| ProviderNvidia | Provider model IDdeepseek-ai/deepseek-v4-pro-0813 | Regionglobal | Input / 1MN/A | Output / 1MN/A | Context1M | Updated |
| ProviderAlibaba Token Plan | Provider model IDdeepseek-v4-pro-0813 | Regionglobal | Input / 1MN/A | Output / 1MN/A | Context1M | Updated |
| ProviderSCNet Token Plan | Provider model IDDeepSeek-V4-Pro-0813 | Regionglobal | Input / 1MN/A | Output / 1MN/A | Context1M | Updated |
| ProviderAlibaba Token Plan (China) | Provider model IDdeepseek-v4-pro-0813 | Regionglobal | Input / 1MN/A | Output / 1MN/A | Context1M | Updated |
| ProviderCrofAI | Provider model IDdeepseek-v4-pro-0813 | Regionglobal | Input / 1M$0.35 | Output / 1M$0.80 | Context1M | Updated |
| ProviderOrcaRouter | Provider model IDdeepseek/deepseek-v4-pro-0813 | Regionglobal | Input / 1M$0.442 | Output / 1M$0.884 | Context1M | Updated |
| ProviderRunInfra | Provider model IDdeepseek-ai/DeepSeek-V4-Pro-0813 | Regionglobal | Input / 1M$0.60 | Output / 1M$1.9 | Context1M | Updated |
| ProviderMerge Gateway | Provider model IDdeepseek/deepseek-v4-pro-0813 | Regionglobal | Input / 1M$0.66 | Output / 1M$1.98 | Context1M | Updated |
| ProviderVercel AI Gateway | Provider model IDdeepseek/deepseek-v4-pro-0813 | Regionglobal | Input / 1M$0.66 | Output / 1M$1.98 | Context1M | Updated |
| ProviderAIHubMix | Provider model IDdeepseek-v4-pro-0813 | Regionglobal | Input / 1M$0.6918 | Output / 1M$2.08 | Context1M | Updated |
| ProviderOpenRouter | Provider model IDdeepseek/deepseek-v4-pro-0813 | Regionglobal | Input / 1M$1.05 | Output / 1M$3.15 | Context1M | Updated |
| ProviderNanoGPT | Provider model IDdeepseek/deepseek-v4-pro-0813 | Regionglobal | Input / 1M$1.1 | Output / 1M$2.5 | Context1M | Updated |
| ProviderDeep Infra | Provider model IDdeepseek-ai/DeepSeek-V4-Pro-0813 | Regionglobal | Input / 1M$1.3 | Output / 1M$2.6 | Context1M | Updated |
| ProviderWeights & Biases | Provider model IDdeepseek-ai/DeepSeek-V4-Pro-0813 | Regionglobal | Input / 1M$1.31 | Output / 1M$3.96 | Context1M | Updated |
| ProviderCloudflare Workers AI | Provider model ID@cf/deepseek-ai/deepseek-v4-pro-0813 | Regionglobal | Input / 1M$1.32 | Output / 1M$3.96 | Context1M | Updated |
| ProviderHugging Face | Provider model IDdeepseek-ai/DeepSeek-V4-Pro-0813 | Regionglobal | Input / 1M$1.32 | Output / 1M$3.96 | Context1M | Updated |
| ProviderDigitalOcean | Provider model IDdeepseek-v4-pro-0813 | Regionglobal | Input / 1M$1.32 | Output / 1M$3.96 | Context1M | Updated |
| ProviderArcee | Provider model IDdeepseek/deepseek-v4-pro-0813 | Regionglobal | Input / 1M$1.32 | Output / 1M$3.96 | Context1M | Updated |
| ProviderTogether AI | Provider model IDdeepseek-ai/DeepSeek-V4-Pro-0813 | Regionglobal | Input / 1M$1.32 | Output / 1M$3.96 | Context1M | Updated |
| ProviderOfox | Provider model IDdeepseek/deepseek-v4-pro-0813 | Regionglobal | Input / 1M$1.32 | Output / 1M$3.96 | Context1M | Updated |
| ProviderRequesty | Provider model IDdeepseek-v4-pro-0813 | Regionglobal | Input / 1M$1.32 | Output / 1M$3.96 | Context1M | Updated |
| ProviderBaseten | Provider model IDdeepseek-ai/DeepSeek-V4-Pro-0813 | Regionglobal | Input / 1M$1.32 | Output / 1M$3.96 | Context1M | Updated |
| ProviderEmpirioLabs AI | Provider model IDdeepseek-v4-pro-0813 | Regionglobal | Input / 1M$1.32 | Output / 1M$3.96 | Context1M | Updated |
| ProviderKilo Gateway | Provider model IDdeepseek/deepseek-v4-pro-0813 | Regionglobal | Input / 1M$1.32 | Output / 1M$3.96 | Context1M | Updated |
| ProviderFireworks AI | Provider model IDaccounts/fireworks/models/deepseek-v4-pro-0813 | Regionglobal | Input / 1M$1.32 | Output / 1M$3.96 | Context1M | Updated |
| ProviderVolcengine Ark | Provider model IDdeepseek-v4-pro-ga-260813 | Regionglobal | Input / 1M$1.34 | Output / 1M$4.01 | Context1M | Updated |
| ProviderVivgrid | Provider model IDdeepseek-v4-pro-0813 | Regionglobal | Input / 1M$1.35 | Output / 1M$3 | Context1M | Updated |
| ProviderCharm Hyper | Provider model IDdeepseek-v4-pro-0813 | Regionglobal | Input / 1M$1.44 | Output / 1M$4.31 | Context1M | Updated |
| ProviderVenice AI | Provider model IDdeepseek-v4-pro-0813 | Regionglobal | Input / 1M$1.65 | Output / 1M$4.95 | Context1M | Updated |
| ProviderCortecs | Provider model IDdeepseek-v4-pro-0813 | Regionglobal | Input / 1M$2 | Output / 1M$4 | Context1M | Updated |
DeepSeek-V4-Pro Runtime Performance
Provider-specific output speed and catalog latency for DeepSeek-V4-Pro-0813. Runtime does not affect the capability score.
Sort by
Provider | Output Speed | Catalog Latency | Max Input | Max Output | Updated |
|---|
| ProviderDeepSeek | Output Speed45.48 tok/s | Catalog Latency2.00 s | Max Input1M | Max Output393.2K | Updated |
| ProviderDeepInfra | Output Speed33.89 tok/s | Catalog Latency2.40 s | Max Input1M | Max Output393.2K | Updated |
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
DeepSeek-V4-Pro Specifications
Technical details for the model's default version.
- Version
- DeepSeek-V4-Pro-0813
- Released
- Aug 13, 2026
- Knowledge cutoff
- Unknown
- Parameters
- 1.6T
- Context window
- 1M
- Max output
- 393.2K
- Inputs
- text
- Outputs
- text
- Open weights
- Yes
- License
- MIT
DeepSeek-V4-Pro Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
Sort by
Version | Released | LLMBoard | Parameters | Context | Max output | Open weights | License |
|---|
| VersionDeepSeek-V4-Pro-0813 | Released | LLMBoard86.78 | Parameters1.6T | Context1M | Max output393.2K | Open weightsYes | LicenseMIT |
| VersionDeepSeek-V4-Pro-Max | Released | LLMBoardN/A | Parameters1.6T | Context1M | Max output393.2K | Open weightsYes | LicenseMIT |
Models similar to DeepSeek-V4-Pro
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
DE
DeepSeek-V4-Flash
DeepSeek
DE
DeepSeek-V3.2
DeepSeek
DE
DeepSeek-R1
DeepSeek
DE
DeepSeek-V3.1
DeepSeek
DE
DeepSeek-V3
DeepSeek
DE
DeepSeek-R1-Distill-Qwen
DeepSeek
What is DeepSeek-V4-Pro?
Key information about DeepSeek-V4-Pro and its available data.
6T-parameter MoE large language model with 49B activated parameters, low, high, and max reasoning effort, a 1M-token context window, and up to 384K output tokens.
It is served through the OpenAI-compatible Chat Completions and Responses APIs under the ID `deepseek-v4-pro`, with open weights available at `deepseek-ai/DeepSeek-V4-Pro-0813` under the MIT license.
Data as of 2026-09-08.
FAQ
Common questions about DeepSeek-V4-Pro.
When was DeepSeek-V4-Pro released?
DeepSeek-V4-Pro's default version was released on Aug 13, 2026.
How much does DeepSeek-V4-Pro cost?
No official standard PAYG price is currently available for DeepSeek-V4-Pro. The lowest tracked third-party offer starts at $0.35 input and $0.80 output via CrofAI.
Who created DeepSeek-V4-Pro?
DeepSeek-V4-Pro was created by DeepSeek.
What is the context window for DeepSeek-V4-Pro?
The default version has a 1M token context window.
Is DeepSeek-V4-Pro open weight?
Yes. The default version is marked as open weight under MIT.
How many API providers offer DeepSeek-V4-Pro?
30 provider offerings are linked to the default version.