← All comparisons

Llama Nemotron Super 49B v1.5 (Non-reasoning) vs DeepSeek R1 Distill Llama 70B

NVIDIA vs DeepSeek — side-by-side benchmark comparison

Llama Nemotron Super 49B v1.5 (Non-reasoning)DeepSeek R1 Distill Llama 70B
Intelligence Index14.616.0
Coding Index10.511.4
Math Index8.053.7
Output speed (tok/s)48.946.8
Blended price ($/1M)$0.17$0.79
Time to first token (s)0.28s0.33s
aime13.7%67.0%
aime 258.0%53.7%
artificial analysis coding index10.5011.40
artificial analysis intelligence index14.6016.00
artificial analysis math index8.0053.70
gpqa48.1%40.2%
hle4.3%6.1%
ifbench32.9%27.6%
lcr22.0%11.0%
livecodebench29.0%26.6%
math 50077.0%93.5%
mmlu pro69.2%79.5%
scicode23.8%31.3%
tau225.1%21.9%
terminalbench hard3.8%1.5%

Benchmark data from Artificial Analysis.