infer gpt2-13G-fp16 837.65 ms per token

[vicwer]

infer gpt2-13G-fp16 model: ./gpt-2 -m ./ggml-model-f16.bin -n 128

main: seed = 1685426504
gpt2_model_load: loading model from './ggml-model-f16.bin'
gpt2_model_load: n_vocab = 50432
gpt2_model_load: n_ctx   = 4096
gpt2_model_load: n_embd  = 5120
gpt2_model_load: n_head  = 40
gpt2_model_load: n_layer = 40
gpt2_model_load: ftype   = 1
gpt2_model_load: qntvr   = 0
gpt2_model_load: ggml tensor size = 224 bytes
gpt2_model_load: ggml ctx size = 31475.43 MB
gpt2_model_load: memory size =  6400.00 MB, n_mem = 163840
gpt2_model_load: model size  = 25075.20 MB
extract_tests_from_file : No test file found.
test_gpt_tokenizer : 0 tests failed out of 0 tests.
main: prompt: 'He'
main: number of tokens in prompt = 1, first 8 tokens: 3876

main: mem per token = 39071352 bytes
main:     load time = 11362.40 ms
main:   sample time =    24.45 ms
main:  predict time = 107219.80 ms / 837.65 ms per token
main:    total time = 119965.73 ms

but llama-7b-fp16 inference speed:./main -m ./hf_llama/7b/ggml-model-f16.bin -n 128

main: build = 583 (7e4ea5b)
main: seed  = 1685426839
llama.cpp: loading model from ./hf_llama/7b/ggml-model-f16.bin
llama_model_load_internal: format     = ggjt v1 (pre #1405)
llama_model_load_internal: n_vocab    = 32000
llama_model_load_internal: n_ctx      = 512
llama_model_load_internal: n_embd     = 4096
llama_model_load_internal: n_mult     = 256
llama_model_load_internal: n_head     = 32
llama_model_load_internal: n_layer    = 32
llama_model_load_internal: n_rot      = 128
llama_model_load_internal: ftype      = 1 (mostly F16)
llama_model_load_internal: n_ff       = 11008
llama_model_load_internal: n_parts    = 1
llama_model_load_internal: model size = 7B
llama_model_load_internal: ggml ctx size =    0.07 MB
llama_model_load_internal: mem required  = 14645.09 MB (+ 1026.00 MB per state)

llama_init_from_file: kv self size  =  256.00 MB

llama_print_timings:        load time =  8700.02 ms
llama_print_timings:      sample time =    73.76 ms /   128 runs   (    0.58 ms per token)
llama_print_timings: prompt eval time =   201.22 ms /     2 tokens (  100.61 ms per token)
llama_print_timings:        eval time = 22030.23 ms /   127 runs   (  173.47 ms per token)
llama_print_timings:       total time = 30836.57 ms

Is the inference speed of these two models normal?