Head-to-head · 5 models compared
DeepSeek vs GPT
DeepSeek and OpenAI between them account for 5 rated models on the OpenRating Index. DeepSeek's strongest entry is DeepSeek V4 Pro 0813 (max) at 53.4, against GPT's GPT-5.6 Sol (max) at 61.2. Both are scored on the same axes, with the same tests, below.
Who wins on each axis
GPT leads all three axes in this comparison.
OpenRating Index
GPT
61.2
GPT: GPT-5.6 Sol (max) · composite 0–100
Output speed
GPT
164 tok/s
GPT: gpt-oss-120b (high) · blended tok/s, fastest model
Cost / task
GPT
$0.048
GPT: GPT-5.6 Luna (max) · lowest model, per benchmark run
Every rated model, side by side
Sorted by OpenRating Index. Every figure is measured under the same standardised workload.
| Model | Provider | Index | Reasoning | Speed | Cost / task | Input / 1M |
|---|---|---|---|---|---|---|
| GPT-5.6 Sol (max) | GPT | 61.2 | 85.0 | 70 tok/s | $1.23 | $2.2 |
| GPT-5.6 Terra (max) | GPT | 56.4 | 76.8 | 86 tok/s | $0.51 | $0.6 |
| DeepSeek V4 Pro 0813 (max) | DeepSeek | 53.4 | 75.2 | 82 tok/s | $0.25 | $0.22 |
| GPT-5.6 Luna (max) | GPT | 52.5 | 70.8 | 149 tok/s | $0.048 | $0.08 |
| gpt-oss-120b (high) | GPT | 44.5 | 61.8 | 164 tok/s | $0.073 | $0.06 |
Put DeepSeek and GPT head to head yourself
Pick any combination of models and compare their full cost breakdown — reasoning, answer generation, cache overhead, and input payload — in the OpenRating comparison tool.
Open the comparison tool →