Every story and comment from the Hacker News API, in a layout borrowed from 2001.
A comment on Generative AI uses more energy on non-English text.
RESULTS J/req = attributed GPU joules per request; ×EN = J/req relative to English; tok/req = output tokens including reasoning; J/tok = total J / total output tokens; Acc S/L = strict/lenient accuracy (%); cap = responses truncated at 5,000 tokens (out of 50).
nw-flash (DeepSeek V4.1 Flash 1017 weights) Language J/req ×EN tok/req J/tok Acc S/L cap English 23.4 1.00 174 0.134 100/100 0
Chinese 16.8 0.72 389 0.043 96/96 0
Russian 18.8 0.80 352 0.054 96/96 0
French 25.1 1.07 272 0.092 98/98 0
S. Pashto 69.2 2.96 752 0.092 92/92 1
Tigrinya 65.3 2.79 2299 0.028 0/76 8
Shan 98.5 4.21 2823 0.035 48/54 14
Tibetan 42.3 1.81 2227 0.019 0/76 9
nw-small (Qwen3.8-27B-FP8) Language J/req ×EN tok/req J/tok Acc S/L cap English 62.2 1.00 300 0.207 100/100 0 Chinese 154.1 2.48 605 0.255 94/94 1 Russian 147.7 2.37 576 0.257 96/96 1 French 163.3 2.62 323 0.506 98/98 0 S. Pashto 777.2 12.49 1510 0.515 84/84 6 Tigrinya 823.8 13.24 2458 0.335 0/76 10 Shan 1407.4 22.62 3422 0.411 48/52 20 Tibetan 1455.5 23.39 3199 0.455 0/56 20
nw-large (Kimi K3.1 Pro) Language J/req ×EN tok/req J/tok Acc S/L cap English 169.5 1.00 391 0.434 100/100 0 Chinese 113.8 0.67 491 0.232 98/98 0 Russian 78.7 0.46 527 0.150 98/98 0 French 298.3 1.76 406 0.735 98/98 0 S. Pashto 119.7 0.71 903 0.133 96/96 0 Tigrinya 1096.6 6.47 1870 0.586 38/86 2 Shan 503.6 2.97 2895 0.174 52/52 12 Tibetan 593.2 3.50 2298 0.258 0/82 4
GROUP COMPARISON (four low-resource versus four high-resource languages) Endpoint Output tokens Energy/request J/output token Strict acc H/L Lenient acc H/L nw-flash 6.83× 3.27× 0.48× 97.5/35.0 97.5/74.5 nw-small 5.87× 8.46× 1.44× 97.0/33.0 97.0/67.0 nw-large 4.39× 3.50× 0.80× 98.5/46.5 98.5/79.0
There are no comments on this comment yet.
RESULTS J/req = attributed GPU joules per request; ×EN = J/req relative to English; tok/req = output tokens including reasoning; J/tok = total J / total output tokens; Acc S/L = strict/lenient accuracy (%); cap = responses truncated at 5,000 tokens (out of 50).
nw-flash (DeepSeek V4.1 Flash 1017 weights) Language J/req ×EN tok/req J/tok Acc S/L cap English 23.4 1.00 174 0.134 100/100 0
Chinese 16.8 0.72 389 0.043 96/96 0
Russian 18.8 0.80 352 0.054 96/96 0
French 25.1 1.07 272 0.092 98/98 0
S. Pashto 69.2 2.96 752 0.092 92/92 1
Tigrinya 65.3 2.79 2299 0.028 0/76 8
Shan 98.5 4.21 2823 0.035 48/54 14
Tibetan 42.3 1.81 2227 0.019 0/76 9
nw-small (Qwen3.8-27B-FP8) Language J/req ×EN tok/req J/tok Acc S/L cap English 62.2 1.00 300 0.207 100/100 0 Chinese 154.1 2.48 605 0.255 94/94 1 Russian 147.7 2.37 576 0.257 96/96 1 French 163.3 2.62 323 0.506 98/98 0 S. Pashto 777.2 12.49 1510 0.515 84/84 6 Tigrinya 823.8 13.24 2458 0.335 0/76 10 Shan 1407.4 22.62 3422 0.411 48/52 20 Tibetan 1455.5 23.39 3199 0.455 0/56 20
nw-large (Kimi K3.1 Pro) Language J/req ×EN tok/req J/tok Acc S/L cap English 169.5 1.00 391 0.434 100/100 0 Chinese 113.8 0.67 491 0.232 98/98 0 Russian 78.7 0.46 527 0.150 98/98 0 French 298.3 1.76 406 0.735 98/98 0 S. Pashto 119.7 0.71 903 0.133 96/96 0 Tigrinya 1096.6 6.47 1870 0.586 38/86 2 Shan 503.6 2.97 2895 0.174 52/52 12 Tibetan 593.2 3.50 2298 0.258 0/82 4
GROUP COMPARISON (four low-resource versus four high-resource languages) Endpoint Output tokens Energy/request J/output token Strict acc H/L Lenient acc H/L nw-flash 6.83× 3.27× 0.48× 97.5/35.0 97.5/74.5 nw-small 5.87× 8.46× 1.44× 97.0/33.0 97.0/67.0 nw-large 4.39× 3.50× 0.80× 98.5/46.5 98.5/79.0