DeepSeek
DeepSeek V4 Pro (Reasoning, High Effort)
Quality Score50.6
Value Score46.0
Reliability44.7
Cache Discount99%
Model specification
- Reasoning
- reasoning
- Input modalities
- text
- Output modalities
- text
- Context window
- 1M tokens
- Weights
- open weights
Token prices USD per 1M tokens
- Input
- $0.435
- Output
- $0.870
- Cached Input
- $0.0036
- Cached Output
- not available
- Cost per task
- $0.043
Capability
- Intelligence
- 43.7
- Coding
- 58.7
- Agentic
- 35.3
- Omniscience
- -10.6
- Correct
- 41.4%
- Blended price
- $0.177
Answer outcomes
Correct 41.4%Incorrect 52.0%Abstained 6.6%
Artificial Analysis benchmarks
- GDPval-AA v2
- 40%
- τ³-Banking
- 26%
- Terminal-Bench v2.1
- 65%
- SciCode
- 46%
- Humanity’s Last Exam
- 35%
- GPQA Diamond
- 91%
- CritPt
- 10%
- AA-Omniscience
- 45%
- AA-LCR
- 67%
Similar models
- DeepSeek V4 Pro (Reasoning, Max Effort)DeepSeekQuality Score 52.0Value Score 47.6
- DeepSeek V4 Flash 0731 (Reasoning, Max Effort)DeepSeekQuality Score 55.1Value Score 50.5
- DeepSeek V4 Pro 0813 (Reasoning, Max Effort)DeepSeekQuality Score 63.3Value Score 47.9
- DeepSeek V4 Pro (Non-reasoning)DeepSeekQuality Score 31.8Value Score 33.5
- DeepSeek V4 Flash (Non-reasoning)DeepSeekQuality Score 25.5Value Score 31.6
- Kimi K2.7 CodeKimiQuality Score 49.2Value Score 38.4