Back to Benchmarks
GPT-5.4 (xhigh) vs DeepSeek R1 Distill Llama 70B
GPT-5.4 (xhigh)
OpenAI
vs
DeepSeek R1 Distill Llama 70B
DeepSeek
via DeepInfra
DeepSeek R1 Distill Llama 70B
(DeepInfra)
wins 6 of 11
Index Scores
57.2
+41.2
Intelligence
16.0
57.3
+45.9
Coding
11.4
0.0
Math
+53.7
53.7
Benchmarks
0.0
MMLU Pro
+79.5
79.5
92.0
+51.8
GPQA
40.2
41.6
+35.5
HLE
6.1
0.0
LiveCode
+26.6
26.6
56.6
+25.4
SciCode
31.2
0.0
MATH-500
+93.5
93.5
0.0
AIME
+67.0
67.0
Speed
32
TPS
+10
42
Shared
Winning delta
crafted by
bart stefanski