DeepSeek
availableShows if the model has enough results for an index.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.
IndexOverall score. 50 is the middle.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
50 is the middle of the board. The range shows the doubt in the index.
CapabilitiesScore per category. 50 is the middle.
50 is the middleResults
32 counted| BenchmarkThe test name. | CategoryThe capability that the test measures. | ResultThe score from the publisher. | IndexThis result on the index scale. | RunThe settings of the run. | DateDate of the result. | Published byThe source of the result. |
|---|---|---|---|---|---|---|
| τ²-Bench Tool-Agent-User Evaluation | Agentic | 96.2% | 67.8 | — | — | Victor Barres et al. |
| Harvard-MIT Mathematics Tournament February 2026 | Math | 95.2% | 61.4 | — | — | Qwen |
| LiveCodeBench Pass@1 with Chain-of-Thought | Coding | 93.5% | — | — | — | DeepSeek |
| Artificial Analysis GPQA Diamond | Knowledge | 92.8% | 63.7 | — | — | Artificial Analysis |
| Apex Shortlist | Math | 90.2% | — | — | — | DeepSeek-AI |
| Graduate-Level Google-Proof Q&A | Knowledge | 90.1% | 61.4 | — | — | David Rein et al. |
| GPQA Diamond | Knowledge | 90.1% | 61.4 | — | — | David Rein et al. |
| IMOAnswerBench | Math | 89.8% | — | — | — | DeepSeek-AI |
| Terminal-Bench 2.1 (provider run) | Agentic | 87.9% | 75.9 | — | — | DeepSeek-AI |
| Terminal-Bench 2.1 (provider run) | Agentic | 87.9% | 75.9 | — | — | DeepSeek-AI |
| Massive Multitask Language Understanding Professional | Knowledge | 87.5% | 58.4 | — | — | Yubo Wang et al. |
| Chinese-SimpleQA | Knowledge | 84.4% | — | — | — | DeepSeek-AI |
| MRCR 1M | Reasoning | 83.5% | — | — | — | DeepSeek-AI |
| BrowseComp | Agentic | 83.4% | 69.8 | — | — | OpenAI |
| CyberGym | Agentic | 83.3% | 73.2 | — | — | Zhun Wang et al. |
| Software Engineering Benchmark Verified | Coding | 80.6% | 62.2 | — | — | Carlos E. Jimenez et al. |
| Artificial Analysis Long Context Reasoning | Reasoning | 80.3% | 63.8 | — | — | Artificial Analysis |
| Artificial Analysis IFBench | Instruction | 76.5% | 66.7 | — | — | Artificial Analysis |
| Toolathlon-Verified | Agentic | 74.1% | 72.3 | — | — | Moonshot AI |
| MCP Atlas | Agentic | 73.6% | 63.6 | — | — | OpenAI |
| DeepSeek DSBench FullStack | Coding | 71.1% | — | — | — | DeepSeek-AI |
| Artificial Analysis Coding Index | Coding | 68.8% | 67.5 | — | — | Artificial Analysis |
| DeepSeek DSBench Hard | Coding | 67.2% | — | — | — | DeepSeek-AI |
| DeepSWE | Agentic | 62.7% | 69.4 | — | — | Datacurve AI |
| CorpusQA 1M | Reasoning | 62.0% | — | — | — | DeepSeek-AI |
| NL2Repo | Coding | 61.5% | 73.4 | — | — | MiniMax |
| Humanity's Last Exam with tools | Agentic | 60.0% | 72.7 | — | — | DeepSeek-AI |
| OpenHarmony Bench v1.0 | Coding | 59.0% | 68.1 | — | — | OpenHarmony Bench authors |
| Measuring Short-Form Factuality in Large Language Models | Knowledge | 57.9% | — | — | — | Jason Wei et al. |
| SWE-bench Pro | Coding | 55.4% | 57.6 | — | — | Xiang Deng et al. |
| GDPval-AA normalized | Agentic | 54.5% | 77.4 | — | — | Artificial Analysis |
| Artificial Analysis Intelligence Index | Knowledge | 53.2% | 88.9 | — | — | Artificial Analysis |
| Toolathlon | Agentic | 51.8% | 66.3 | — | — | OpenAI |
| Artificial Analysis SciCode | Coding | 51.0% | 63.6 | — | — | Artificial Analysis |
| Artificial Analysis EnterpriseOps-Gym | Agentic | 49.6% | 75.2 | — | — | Artificial Analysis |
| Artificial Analysis Agentic Index | Agentic | 49.6% | 77.6 | — | — | Artificial Analysis |
| Artificial Analysis Omniscience Accuracy | Knowledge | 49.1% | 74.6 | — | — | Artificial Analysis |
| Humanity's Last Exam | Knowledge | 42.7% | 65.0 | — | — | Center for AI Safety et al. |
| Artificial Analysis Humanity's Last Exam | Knowledge | 41.0% | 69.0 | — | — | Artificial Analysis |
| Apex | Math | 38.3% | — | — | — | DeepSeek-AI |
| AutomationBench | Agentic | 31.8% | 68.4 | — | — | Moonshot AI |
| Agents' Last Exam | Agentic | 25.7% | 65.4 | — | — | DeepSeek-AI |
| APEX-Agents-AA | Agentic | 24.3% | 60.3 | — | — | Artificial Analysis / Mercor |
| Critical Physics Tasks | Reasoning | 18.0% | 76.1 | — | — | Artificial Analysis |
32 benchmarks count, from 34 of 44 results. A grey row does not count. Too few models took that benchmark.