Model AI terbaik untuk instruksi
Peringkat model berdasar kepatuhan instruksi. Diperbarui Agustus 2026. Berdasar data benchmark independen.
| # | Model | Penyedia | Skor | Kecepatan | Harga (output) |
|---|---|---|---|---|---|
| 1 | Claude Mythos 5 | Anthropic | 83.0 | 50 t/s | $50/M |
| 2 | Claude Sonnet 4.6 (max) | Anthropic | 78.0 | 50 t/s | $15/M |
| 3 | GLM-5.1 | Z.ai | 77.0 | 74 t/s | $4.4/M |
| 4 | MiniMax-M2.7 | MiniMax | 74.0 | 40 t/s | $2/M |
| 5 | Claude Fable 5.1 | Anthropic | 60.0 | 85 t/s | $15/M |
| 6 | Gemini 3.8 Flash | 60.0 | 302 t/s | $3.75/M | |
| 7 | Gemini 3.8 Flash (Medium Reasoning) | 60.0 | 315 t/s | $3.75/M | |
| 8 | Gemini 3.8 Flash (Low Reasoning) | 60.0 | 340 t/s | $3.75/M | |
| 9 | Mistral Medium 3.5 | Mistral | 59.0 | 60 t/s | $6/M |
| 10 | Claude 4.5 Haiku | Anthropic | 56.0 | 80 t/s | $15/M |
| 11 | GPT-5.5 Pro | OpenAI | 55.0 | 60 t/s | $10/M |
| 12 | Nova 2.0 Pro Preview | Amazon | 54.0 | 40 t/s | $3.2/M |
| 13 | GPT-oss-120B | OpenAI | 50.0 | 240 t/s | $0.4/M |
| 14 | Claude Fable 5.1 (XHigh) | Anthropic | 45.0 | 78 t/s | $15/M |
| 15 | GPT-6 Astra (xhigh) | OpenAI | 45.0 | 76 t/s | $50/M |
| 16 | Claude Fable 5.1 (High) | Anthropic | 45.0 | 92 t/s | $15/M |
| 17 | GPT-6 Astra (high) | OpenAI | 45.0 | 90 t/s | $50/M |
| 18 | GPT-6 Luna | OpenAI | 45.0 | 120 t/s | $20/M |
| 19 | Claude Opus 5.5 (XHigh) | Anthropic | 45.0 | 62 t/s | $25/M |
| 20 | GPT-6 Luna (xhigh) | OpenAI | 45.0 | 118 t/s | $20/M |
Claude Mythos 5
GLM-5.1
MiniMax-M2.7
Gemini 3.8 Flash
Mistral Medium 3.5
GPT-5.5 Pro
Nova 2.0 Pro PreviewTop 5 \u2014 Perbandingan cepat
Metodologi
Peringkat berdasar evaluasi benchmark independen. Skor dihitung dari beberapa benchmark terstandar dan dinormalisasi ke 0–100.