AA 30 Oct 評估更新:DeepSeek V4.1 Flash (Non-reasoning) 首評 19、Nemotron 3.5 Super VL 降價;Gemini 4 Carbon 59.4 仍居首
Artificial Analysis 公布之 Intelligence Index v5、median 輸出 tps、e2e 與牌價(第三方評測)
資料日 2026-10-31(AA changelog 最新條目 30 Oct 2026)來源 artificialanalysis.ai已取得 v5 智力分的模型數 86(另 82 個模型僅有效能數據、智力欄空白)🕒 產出 2026-10-31 08:00 CST
最高 Intelligence Index v5
Gemini 4 Carbon (High)|59.4
最高 median 輸出速度
Celeris-1|1663.0 tps
DeepSeek V4.1 Flash 1017 (Max)
32.5|175.1 tps|e2e 15.37s
DeepSeek V4.1 Flash 1017 (Non-reasoning)
18.5|1486.0 tps|e2e 1.49s
📊 AA 公布數字(median)
tps=輸出 tokens/s;e2e=單任務端到端回應時間;價格為每 1M tokens 牌價。智力為 Artificial Analysis Intelligence Index v5.0。
模型智力 v5輸出 tpse2e價格 /1M(in/out)
Gemini 4 Carbon (High)59.484.3123.43s$4.00 / $20.00
Claude Opus 5.5 (Max)58.496.6668.83s$4.00 / $20.00
Claude Opus 5.5 (Xhigh)57.085.4127.05s$4.00 / $20.00
Claude Haiku 5.5 (Max)36.2236.5338.79s$0.10 / $0.50
Claude Haiku 5.5 (Xhigh)34.3186.471.71s$0.10 / $0.50
DeepSeek V4.1 Flash 1017 (Max)32.5175.115.37s$0.30 / $1.20
GLM-5.3 (Max)31.777.634.00s$1.40 / $4.40
Claude Haiku 5.5 (High)29.6172.720.87s$0.10 / $0.50
Step 5 Preview28.5165.416.03s$1.00 / $2.70
Mistral Large 4 (High)26.0161.316.92s$0.70 / $2.10
Mistral Large 4 (Medium)23.1未公布未公布$0.70 / $2.10
Claude Haiku 5.5 (Medium)24.6169.413.75s$0.10 / $0.50
DeepSeek V4 Pro 0813 (Max)23.180.732.63s$1.32 / $3.96
DeepSeek V4.1 Flash 1017 (Non-reasoning)18.51486.01.49s$0.45 / $1.80
K2 Horizon 375B A23B17.8362.09.22s$0 / $0
Nemotron 3.5 Super VL13.6474.15.96s$0.20 / $0.50
Celeris-1未公布1663.00.87s$0.20 / $0.70