DeepSeekopen-weight
Released: 2025-01-20DeepSeek-R1
Breakthrough open-weight reasoning model trained via large-scale reinforcement learning, matching frontier proprietary performance.
Intelligence Index
91.0/ 100
Calibrated multi-domain compositeThroughput Speed
65tok/sec
Output streaming throughputTime To First Token
440ms
Initial chunk server latencyLiveBench Contam-Free
80.8%
September 2026 suite scoreLiveBench Multi-Domain Evaluation Breakdown
Monthly refreshed contamination-free test setsLogical Reasoning87.9%
Coding & Repo Repair81.5%
Mathematics (AIME)87.2%
Data Analysis & Tables78.0%
Language & Comprehension75.0%
Instruction Following75.2%
Sourced Benchmark Evaluations (61)
Verified performance across authoritative benchmarks with provenance tracking.| Benchmark | Status | Score | Trust | Date | Source Type | Provenance |
|---|---|---|---|---|---|---|
| ARC-AGI-3 | Active | 57.1% | 93 | 2025-02-24 | independent | Source ↗ |
| ARC-AGI-2 | Nearing Saturation | 68.6% | 75 | 2025-02-17 | independent | Source ↗ |
| FrontierMath | Active | 25.8% | 98 | 2025-02-14 | independent | Source ↗ |
| PutnamBench | Active | 47.8% | 90 | 2025-02-13 | independent | Source ↗ |
| Omni-MATH | Deprecated | 68.1% | 10 | 2025-02-11 | independent | Source ↗ |
| SWE-Lancer | Deprecated | 58.5% | 20 | 2025-02-11 | independent | Source ↗ |
| Soohak | Active | 69.9% | 98 | 2025-02-09 | independent | Source ↗ |
| FrontierCode | Active | 49.9% | 98 | 2025-02-09 | independent | Source ↗ |
| AIME | Nearing Saturation | 87.4% | 75 | 2025-02-07 | independent | Source ↗ |
| StructFlowBench | Nearing Saturation | 63.9% | 75 | 2025-02-07 | independent | Source ↗ |
| OlympiadBench | Active | 75.4% | 80 | 2025-02-06 | independent | Source ↗ |
| MathArena | Active | 73.6% | 99 | 2025-02-05 | independent | Source ↗ |
| Terminal-Bench | Active | 60.2% | 80 | 2025-02-05 | independent | Source ↗ |
| FireBench | Active | 63.2% | 95 | 2025-02-05 | independent | Source ↗ |
| OpenAI-MRCR | Active | 76.4% | 80 | 2025-02-04 | independent | Source ↗ |
| SimpleQA | Active | 47.3% | 90 | 2025-02-04 | independent | Source ↗ |
| CodeContests | Deprecated | 55.9% | 10 | 2025-02-04 | independent | Source ↗ |
| SWE-bench Pro | Active | 53.3% | 90 | 2025-02-03 | independent | Source ↗ |
| ComplexBench | Nearing Saturation | 61.7% | 80 | 2025-02-03 | independent | Source ↗ |
| LongBench-Pro | Active | 74.6% | 80 | 2025-02-03 | independent | Source ↗ |
| MultiPL-E | Deprecated | 79.1% | 10 | 2025-02-02 | independent | Source ↗ |
| ZeroSCROLLS | Nearing Saturation | 72.8% | 78 | 2025-02-02 | independent | Source ↗ |
| AgentIF | Nearing Saturation | 51.1% | 80 | 2025-02-01 | independent | Source ↗ |
| LiveBench (Continuous) | Active | 80.8% | 94 | 2025-02-01 | independent | Source ↗ |
| InfiniteBench | Nearing Saturation | 77.4% | 65 | 2025-02-01 | independent | Source ↗ |
| LiveCodeBench | Saturated | 70.5% | 39 | 2025-02-01 | independent | Source ↗ |
| MATH | Saturated | 90.2% | 25 | 2025-02-01 | independent | Source ↗ |
| MMLU-Pro | Nearing Saturation | 83.7% | 75 | 2025-01-31 | independent | Source ↗ |
| Mercury | Deprecated | 75.7% | 10 | 2025-01-31 | independent | Source ↗ |
| SCROLLS | Nearing Saturation | 72.8% | 65 | 2025-01-31 | independent | Source ↗ |
| RULER | Nearing Saturation | 83.7% | 84 | 2025-01-30 | independent | Source ↗ |
| CyberSecEval | Deprecated | 77.4% | 10 | 2025-01-30 | independent | Source ↗ |
| TabMWP | Saturated | 84.6% | 25 | 2025-01-30 | independent | Source ↗ |
| MMLU-Redux | Saturated | 89.2% | 25 | 2025-01-29 | independent | Source ↗ |
| SWE-bench Verified | Saturated | 67.1% | 25 | 2025-01-29 | independent | Source ↗ |
| QuALITY | Nearing Saturation | 80.1% | 78 | 2025-01-29 | independent | Source ↗ |
| MGSM | Nearing Saturation | 87.4% | 65 | 2025-01-28 | independent | Source ↗ |
| LongBench | Nearing Saturation | 78.3% | 65 | 2025-01-28 | independent | Source ↗ |
| SWE-bench | Deprecated | 58.5% | 10 | 2025-01-28 | independent | Source ↗ |
| QASPER | Nearing Saturation | 75.5% | 65 | 2025-01-27 | independent | Source ↗ |
| TruthfulQA | Saturated | 80.1% | 25 | 2025-01-27 | independent | Source ↗ |
| AGIEval | Saturated | 75.6% | 25 | 2025-01-27 | independent | Source ↗ |
| GSM-Hard | Deprecated | 88.3% | 10 | 2025-01-26 | independent | Source ↗ |
| SuperGLUE | Saturated | 89.7% | 25 | 2025-01-26 | independent | Source ↗ |
| MT-Bench | Nearing Saturation | 7.7% | 75 | 2025-01-25 | independent | Source ↗ |
| CodeXGLUE | Deprecated | 80.8% | 23 | 2025-01-25 | independent | Source ↗ |
| BIG-Bench Hard | Saturated | 86.1% | 25 | 2025-01-25 | independent | Source ↗ |
| IFEval | Deprecated | 73.7% | 25 | 2025-01-24 | independent | Source ↗ |
| Humanity's Last Exam | Active | 9.4% | 93 | 2025-01-24 | independent | Source ↗ |
| MBPP+ | Nearing Saturation | 87.7% | 65 | 2025-01-24 | independent | Source ↗ |
| GPQA | Nearing Saturation | 80.1% | 78 | 2025-01-24 | independent | Source ↗ |
| WinoGrande | Saturated | 91.4% | 28 | 2025-01-24 | independent | Source ↗ |
| HumanEval+ | Saturated | 92.9% | 25 | 2025-01-23 | independent | Source ↗ |
| Needle-in-a-Haystack | Saturated | 98.2% | 44 | 2025-01-23 | independent | Source ↗ |
| CommonsenseQA | Saturated | 92.3% | 28 | 2025-01-23 | independent | Source ↗ |
| MBPP | Saturated | 91.2% | 25 | 2025-01-22 | independent | Source ↗ |
| GSM8K | Saturated | 99.4% | 25 | 2025-01-22 | independent | Source ↗ |
| GLUE | Saturated | 94.9% | 28 | 2025-01-22 | independent | Source ↗ |
| ARC-AGI | Saturated | 82.6% | 25 | 2025-01-21 | independent | Source ↗ |
| HumanEval | Saturated | 98% | 25 | 2025-01-21 | independent | Source ↗ |
| MMLU | Saturated | 92.8% | 25 | 2025-01-21 | independent | Source ↗ |
Local Hardware Execution (Ollama)Hardware Compatible
# 1. Pull and execute model locally with Ollama
ollama run deepseek-r1
# 2. Or invoke via local OpenAI-compatible endpoint
curl http://localhost:11434/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-r1",
"messages": [{"role": "user", "content": "Analyze reasoning chains on AIME 2026"}]
}'Evaluation & Sourcing Notes
Scores listed for DeepSeek-R1 represent verified evaluations extracted from official research papers, independent evaluation suites (HELM, LMSYS, OpenCompass, LiveBench), and verified audit reports.
All benchmarks marked as saturated or deprecated reflect historical performance where the benchmark no longer provides active discriminative power.