DeepSeek-R1
Classic open reasoning model for transparent math, coding, and deliberate problem solving
Shogo Agent Index
Estimated#87 of 9221.1range 13.5–28.6 · 15% of weight measured
Agentic coding
21
estimated
Long-horizon autonomy
31
50% measured
Knowledge-work agents
18
20% measured
Computer use
10
estimated
Security & context learning
21
estimated
Price
- Input
- $0.50/M
- Output
- $4.88/M
- Cache read
- —/M
- 3:1 blend
- $1.59/M
- Agent-effective
- $0.71/M
Agent-effective = 20 input tokens per output token, 85% cache hits. Why.
Agentic benchmarks
| Benchmark | Score | Normalised | Index weight | Harness / source |
|---|---|---|---|---|
| SWE-bench Verified | est. | 8% | predicted 56 | |
| Terminal-Bench 2.0 | est. | 8% | predicted 15 | |
| DeepSWE | est. | 6% | predicted 13 | |
| FrontierSWE | est. | 5% | predicted 1 | |
| CursorBench | est. | 4% | predicted 20 | |
| GSO-Bench | est. | 4% | predicted 3 | |
| METR Time Horizons | 53.8% | 10% | source | |
| Vending-Bench 2 | est. | 6% | predicted 4 | |
| PostTrainBench | est. | 4% | predicted 13 | |
| APEX-Agents | est. | 8% | predicted 15 | |
| GDPval | est. | 7% | predicted 21 | |
| Remote Labor Index | est. | 5% | predicted 0 | |
| DeepResearch Bench | 35.1% | 5% | source | |
| OSWorld 2.0 | est. | 6% | predicted 1 | |
| OSWorld | est. | 4% | predicted 24 | |
| Cybench | est. | 5% | predicted 28 | |
| CL-bench | est. | 5% | predicted 14 |
Other benchmarks
| Benchmark | Category | Score | Source |
|---|---|---|---|
| Aider Polyglot | Coding | 71.4% | source |
| AlgoTune | Coding | 1.70× | source |
| ARC-AGI | Reasoning | 21.2% | source |
| ARC-AGI-2 | Reasoning | 1.3% | source |
| BALROG | Long-horizon | 34.9% | source |
| CritPt | Reasoning | 1.1% | source |
| Fiction.LiveBench | Long context | 75.0% | source |
| GPQA Diamond | Reasoning | 76.3% | source |
| MATH Level 5 | Math | 96.6% | source |
| OTIS Mock AIME | Math | 66.4% | source |
| SciCode | Coding | 38.3% | source |
| SimpleBench | Reasoning | 40.8% | source |