Rankings / DeepSeek
Deepseek R1 Distill Llama 70B
released 2025-01-20
BenchAtlas Index
as of 2026-07-14
base configuration
rank #397
12 families · 5 categories · high
Benchmark evidence
16 results
agentic coding
| τ²-Bench | 21.9 % | independent | Artificial Analysis |
coding
| LiveCodeBench v6 implementation=artificial-analysis | 26.6 % | independent | Artificial Analysis | |
| SciCode | 31.3 % | independent | Artificial Analysis |
external indices
| Intelligence Index v4.1 | 9.9 points | independent | Artificial Analysis | |
| AA Math Index | 53.7 points | independent | Artificial Analysis |
knowledge science
| GPQA Diamond | 55.7 % | independent | Epoch AI Benchmarking Hub 2025-03-10 | |
| GPQA Diamond implementation=artificial-analysis | 40.2 % | independent | Artificial Analysis | |
| Humanity's Last Exam implementation=artificial-analysis | 6.1 % | independent | Artificial Analysis | |
| MMLU-Pro implementation=artificial-analysis | 79.5 % | independent | Artificial Analysis |
long context instruction
| AA-LCR | 11.0 % | independent | Artificial Analysis | |
| IFBench | 27.6 % | independent | Artificial Analysis |
reasoning math
| AIME implementation=artificial-analysis | 67.0 % | independent | Artificial Analysis | |
| AIME year=2025 · implementation=artificial-analysis | 53.7 % | independent | Artificial Analysis | |
| MATH-500 implementation=artificial-analysis | 93.5 % | independent | Artificial Analysis | |
| MATH Level 5 | 89.9 % | independent | Epoch AI Benchmarking Hub 2025-03-10 | |
| OTIS Mock AIME 2024–2025 | 51.4 % | independent | Epoch AI Benchmarking Hub 2025-03-07 |
Agent + model results
systems, not bare-model scores
| agent + model Artificial Analysis harness + Deepseek R1 Distill Llama 70B | Terminal-Bench Hard | 1.5 % | independent | Artificial Analysis |
These scores measure the whole agent system (scaffold, tools, budgets) — they are never merged into the bare model’s numbers.