AI modelleri ve benchmark karşılaştırması
Önde gelen yapay zekâ modellerinin bağımsız testlerdeki genel zekâ puanları. Her firmanın en güçlü ayarı gösteriliyor.
| # | Model | Zekâ endeksi | Bağlam |
|---|---|---|---|
| 1 | Claude Opus 5.5 (max)Anthropic | 58 | 1M |
| 2 | Claude Sonnet 5.5 (max)Anthropic | 56 | 1M |
| 3 | Claude Fable 5.1 (max)Anthropic | 53 | 1M |
| 3 | GPT-6 Astra (max)OpenAI | 53 | 1M |
| 3 | Gemini 4 Argon (high)Google | 53 | 1M |
| 6 | GPT-6.1 Sol (max)OpenAI | 52 | 1M |
| 7 | Muse Spark 1.3 (max)Meta | 48 | 1M |
| 8 | Grok 4.7 (xhigh)SpaceXAI | 46 | 500k |
| 8 | MiMo-V2.6-ProXiaomi | 46 | 1M |
| 10 | Qwen3.8 Max (0902)Alibaba | 45 | 984k |
| 10 | GLM-5.3 (max)Z AI | 45 | 1M |
| 12 | Kimi K3 (max)Kimi | 44 | 1.05M |
| 12 | Step 5 PreviewStepFun | 44 | 1M |
| 14 | Claude Haiku 5.5 (max)Anthropic | 43 | 1M |
| 15 | Gemini 3.8 Flash (high)Google | 41 | 1M |
| 16 | DeepSeek V4.1 Flash (max)DeepSeek | 39 | 1M |
| 17 | MiniMax-M3MiniMax | 29 | 1M |
| 18 | Nemotron 3 UltraNVIDIA | 23 | 262k |
| 19 | Mistral Medium 3.5Mistral | 14 | 256k |
| 19 | Nova 2.0 Pro Preview (medium)Amazon | 14 | 256k |
| 21 | gpt-oss-120b (high)OpenAI | 12 | 131k |
Zekâ endeksi, Artificial Analysis’in akıl yürütme, bilgi, matematik ve kodlama testlerini birleştiren bağımsız puanıdır (yüksek = daha iyi). Bağlam: modelin tek seferde okuyabildiği metin uzunluğu (token).
Kaynak: Artificial Analysis · Güncelleme: 12 Ekim 2026
Kısa yorum
- En yüksek puan: Claude Opus 5.5 (max) (Anthropic) — 58
- Fiyat/performans lideri: MiMo-V2.6-Pro — 46 / $0,13
- En hızlı: gpt-oss-120b (high) — 241 token/s
Maliyet ve hız dahil tam tablo
| Model | Firma | Zekâ endeksi | Bağlam | Görev başı maliyet | Hız (token/sn) |
|---|---|---|---|---|---|
| Claude Opus 5.5 (max) | Anthropic | 58 | 1M | $5,98 | 96 |
| Claude Sonnet 5.5 (max) | Anthropic | 56 | 1M | $5,46 | 137 |
| Claude Fable 5.1 (max) | Anthropic | 53 | 1M | $7,63 | 66 |
| GPT-6 Astra (max) | OpenAI | 53 | 1M | $3,26 | 63 |
| Gemini 4 Argon (high) | 53 | 1M | $1,99 | — | |
| GPT-6.1 Sol (max) | OpenAI | 52 | 1M | $0,72 | 66 |
| Muse Spark 1.3 (max) | Meta | 48 | 1M | $1,60 | 121 |
| Grok 4.7 (xhigh) | SpaceXAI | 46 | 500k | $3,74 | 83 |
| MiMo-V2.6-Pro | Xiaomi | 46 | 1M | $0,13 | 48 |
| Qwen3.8 Max (0902) | Alibaba | 45 | 984k | $5,41 | 36 |
| GLM-5.3 (max) | Z AI | 45 | 1M | $2,01 | 66 |
| Kimi K3 (max) | Kimi | 44 | 1.05M | $2,00 | 41 |
| Step 5 Preview | StepFun | 44 | 1M | $1,03 | 89 |
| Claude Haiku 5.5 (max) | Anthropic | 43 | 1M | $0,21 | 235 |
| Gemini 3.8 Flash (high) | 41 | 1M | $1,24 | 132 | |
| DeepSeek V4.1 Flash (max) | DeepSeek | 39 | 1M | $0,27 | 208 |
| MiniMax-M3 | MiniMax | 29 | 1M | $0,51 | 110 |
| Nemotron 3 Ultra | NVIDIA | 23 | 262k | $0,60 | 155 |
| Mistral Medium 3.5 | Mistral | 14 | 256k | $0,50 | 164 |
| Nova 2.0 Pro Preview (medium) | Amazon | 14 | 256k | — | 122 |
| gpt-oss-120b (high) | OpenAI | 12 | 131k | $0,11 | 241 |
Diğer akıl yürütme ayarları
| Model | Firma | Zekâ endeksi | Bağlam | Görev başı maliyet | Hız (token/sn) |
|---|---|---|---|---|---|
| Claude Opus 5.5 (xhigh) | Anthropic | 56 | 1M | $3,46 | 81 |
| Claude Opus 5.5 (high) | Anthropic | 54 | 1M | $1,82 | 82 |
| Claude Opus 5.5 (medium) | Anthropic | 51 | 1M | $1,34 | 82 |
| Claude Fable 5.1 (xhigh) | Anthropic | 53 | 1M | $5,98 | 62 |
| Claude Fable 5.1 (high) | Anthropic | 51 | 1M | $3,91 | 56 |
| Claude Fable 5.1 (medium) | Anthropic | 49 | 1M | $2,98 | 56 |
| Claude Fable 5.1 (low) | Anthropic | 47 | 1M | $2,37 | 57 |
| Claude Sonnet 5.5 (xhigh) | Anthropic | 52 | 1M | $2,01 | 106 |
| Claude Sonnet 5.5 (high) | Anthropic | 47 | 1M | $0,88 | 105 |
| Claude Sonnet 5.5 (medium) | Anthropic | 41 | 1M | $0,48 | 102 |
| GPT-6 Astra (xhigh) | OpenAI | 52 | 1M | $2,31 | 61 |
| GPT-6 Astra (high) | OpenAI | 51 | 1M | $1,73 | 56 |
| GPT-6 Astra (medium) | OpenAI | 50 | 1M | $1,54 | 55 |
| GPT-6 Astra (low) | OpenAI | 46 | 1M | $0,82 | 54 |
| Muse Spark 1.3 (xhigh) | Meta | 45 | 1M | $1,37 | 259 |
| GPT-6.1 Sol (xhigh) | OpenAI | 51 | 1M | $0,39 | 61 |
| GPT-6.1 Sol (high) | OpenAI | 50 | 1M | $0,32 | 58 |
| GPT-6.1 Sol (medium) | OpenAI | 48 | 1M | $0,21 | 55 |
| GPT-6.1 Sol (low) | OpenAI | 42 | 1M | $0,13 | 53 |
| Grok 4.7 (high) | SpaceXAI | 46 | 500k | $2,73 | 82 |
| Claude Haiku 5.5 (xhigh) | Anthropic | 41 | 1M | $0,12 | 188 |
| Claude Haiku 5.5 (high) | Anthropic | 38 | 1M | $0,08 | 179 |
Tek tek testlerde öne çıkanlar
| Test | Model | Puan |
|---|---|---|
| Humanity's Last Exam | Claude Opus 5.5 (max) | %61,4 |
| Humanity's Last Exam | Claude Fable 5.1 (max) | %59,1 |
| GPQA Diamond | GPT-6 Astra (xhigh) | %96,3 |
| GPQA Diamond | GPT-6 Astra (max) | %96,1 |
| GPQA Diamond | Gemini 3.8 Flash (high) | %95,3 |
Kaynak: Artificial Analysis · Güncelleme: 12 Ekim 2026