Sari la conținut

Română · beta. Traducerea a fost făcută automat și este încă în curs de revizuire. Citește pagina în engleză

Benchmark-uri

Surse
8
Intrări urmărite
22.310
Ultima sincronizare
43 min
Toate sursele se sincronizează normal
  • Arena · Textacum 3 ore
  • Arena · WebDevacum 3 ore
  • Arena · Visionacum 3 ore
  • Arena · Agentacum 3 ore
  • GPQA Diamondacum 43 de minute
  • SWE-benchacum 43 de minute
  • Terminal-Benchacum 43 de minute
  • ARC-AGI-2acum 43 de minute

gemini-4-argon

96.4%±1.2%

95.2%CI97.6%

Google DeepMindoct. 2026oscilație #1–#10

gpt-6-astra

95.8%±1.4%

94.4%CI97.1%

OpenAIsept. 2026oscilație #1–#18

GPQA Diamond ×1221 din 221 modele
ModelOrganizație± CIOscilațieAdaugă la comparație
1gemini-4-argonGoogle DeepMindoct. 2026$2 / $1096.4%95.2%–97.6%#1–10
2gpt-6-astraOpenAIsept. 2026$10 / $5095.8%94.4%–97.1%#1–18
3claude-sonnet-5-5Anthropicoct. 2026$2 / $1095.6%94.2%–97.0%#4–26
4gemini-3.8-flashGoogle DeepMindoct. 2026$0.83 / $4.1395.4%94.0%–96.8%#2–29
5gpt-6.1-solOpenAIoct. 2026$2 / $1095.4%94.0%–96.8%#1–15
6gemini-3.7-flashGoogle DeepMindoct. 2026$0.75 / $3.7594.8%93.5%–96.2%#9–57
7gpt-5.4-pro-2026-03-05OpenAImar. 2026$30 / $18094.6%93.0%–96.2%#20–171
8gemini-3.1-pro-previewGoogle DeepMindsept. 2026$2 / $1294.4%92.8%–96.1%#9–36
9gpt-6-solOpenAIoct. 2026$2 / $1094.3%92.8%–95.8%#7–53
10gemini-3.6-flashGoogle DeepMindsept. 2026$0.75 / $3.7594.1%92.7%–95.5%#16–95
11gpt-5.5-pre-releaseOpenAI——94.0%92.5%–95.5%#22–36
12grok-4.6xAIsept. 2026$2 / $694.0%92.6%–95.5%#15–52
13gpt-5.5-pro-pre-releaseOpenAI——93.9%92.3%–95.5%#23–60
14claude-opus-5Anthropicsept. 2026$5 / $2593.9%92.4%–95.4%#5–11
15gpt-5.6-solOpenAIsept. 2026$4 / $2093.5%91.9%–95.1%#6–25
16grok-4.5xAIsept. 2026$2 / $693.4%92.0%–94.9%#29–70
17gpt-5.6-terraOpenAIsept. 2026$2 / $1293.3%91.8%–94.8%#13–59
18gpt-5.4-2026-03-05OpenAIsept. 2026$2.50 / $1593.3%91.5%–95.1%#12–29
19kimi-k3Moonshotiun. 2026$0.25 / $11.9493.1%91.6%–94.6%#10–39
20gemini-3.5-flashGoogle DeepMindsept. 2026$1.50 / $992.8%91.2%–94.4%#10–26
21grok-4.7xAIoct. 2026$2 / $692.7%91.2%–94.3%#21–72
22qwen3.8-maxAlibabaoct. 2026$2 / $692.7%91.0%–94.4%#12–40
23gemini-3-pro-previewGoogle DeepMind——92.6%91.0%–94.3%#19–120
24qwen3.8-max-0902Alibabaoct. 2026$2 / $692.3%90.6%–94.0%#23–155
25glm-5.2Z.ai (Zhipu AI)iun. 2026$0.62 / $2.4291.9%90.2%–93.5%#9–49
26deepseek-v4-pro-0813DeepSeekaug. 2026$0.66 / $1.9891.7%90.1%–93.2%#40–130
27gpt-5.6-lunaOpenAIsept. 2026$0.20 / $1.2091.6%89.9%–93.3%#30–146
28gpt-5.2-2025-12-11OpenAIdec. 2025$1.75 / $1491.4%89.6%–93.2%#34–92
29claude-opus-4-8Anthropicsept. 2026$5 / $2591.0%89.1%–93.0%#15–33
30deepseek-v4-flash-0731DeepSeekiul. 2026$0.01 / $0.4591.0%89.3%–92.8%#45–244
31deepseek-v4-proDeepSeekaug. 2026$0.66 / $1.9890.9%88.9%–93.0%#17–54
32glm-5.3Z.ai (Zhipu AI)aug. 2026$0.08 / $1.4090.9%89.3%–92.5%#18–52
33MiniMax-M3MiniMaxiun. 2026$0.30 / $1.2090.9%88.9%–93.0%#50–264
34qwen3.7-maxAlibabamai 2026$1.48 / $4.4290.9%88.9%–93.0%#41–128
35kimi-k2.6Moonshotapr. 2026$0.65 / $3.4190.8%89.1%–92.5%#29–78

Se afișează GPQA Diamond exact așa cum a publicat-o sursa — propriul scor și interval de încredere, fără mix. Selectează o a doua axă ca să construiești un scor compus ponderat.

Puținele lucruri din AI și cuantic care au contat cu adevărat săptămâna asta — rezumate cinstit.

Un e-mail pe săptămână. Dezabonare dintr-un clic. Fără urmărire, fără partajare.