🔭 Peringkat: Riset
Literatur ilmiah dan tugas ahli: HLE, GPQA Diamond, dan MMLU-Pro.
🏆 Keseluruhan673💻 Coding392📐 Matematika32🧠 Penalaran391🤖 Agen31📚 Pengetahuan388👁️ Multimodal370🛠️ Agen coding40🔬 Penalaran ilmiah45🌍 Multibahasa12✍️ Menulis53💬 Chat & asisten53🔭 Riset48🔧 Pemanggilan alat31📜 Konteks panjang55⚡ Kecepatan378💰 Nilai377🔓 Bobot terbuka153
Model: 48
| # | Model | Podium Score | Kecepatan |
|---|---|---|---|
| 01 | Seed 2.0 Pro ByteDance | 88.9 | 40 t/s |
| 02 | GPT-5.1 OpenAI | 88.1 | 60 t/s |
| 03 | Claude Opus 4.5 Anthropic | 87.0 | 50 t/s |
| 4 | ERNIE 5.1 Baidu | 87.0 | 60 t/s |
| 5 | Qwen3.5 122B-A10B Alibaba · Terbuka | 86.7 | 126 t/s |
| 6 | GPT-5 OpenAI | 85.7 | 60 t/s |
| 7 | Gemma 4 31B Google · Terbuka | 85.2 | 35 t/s |
| 8 | DeepSeek V3.2 DeepSeek · Terbuka | 85.0 | 60 t/s |
| 9 | Claude Mythos Preview Anthropic · preview | 79.7 | 80 t/s |
| 10 | Claude Opus 4.8 Anthropic | 75.8 | 50 t/s |
| 11 | Gemini 3 Pro Google | 75.8 | 120 t/s |
| 12 | DeepSeek V4 Pro DeepSeek · Terbuka | 75.3 | 66 t/s |
| 13 | Kimi K3 Moonshot AI · Terbuka | 74.8 | 37 t/s |
| 14 | Claude Opus 4.7 Anthropic | 74.5 | 80 t/s |
| 15 | Qwen3.7 Max Alibaba | 74.5 | 203 t/s |
| 16 | Muse Spark Meta | 74.0 | 80 t/s |
| 17 | DeepSeek V4 Flash DeepSeek · Terbuka | 73.1 | 113 t/s |
| 18 | GLM-5.2 Z.ai · Terbuka | 73.0 | 164 t/s |
| 19 | Claude Fable 5 Anthropic | 72.9 | 71 t/s |
| 20 | Claude Opus 5 Anthropic | 72.9 | 60 t/s |
| 21 | Gemini 3.1 Pro Google | 72.9 | 136 t/s |
| 22 | GPT-5.5 OpenAI | 72.9 | 60 t/s |
| 23 | Claude Opus 4.6 Anthropic | 72.2 | 50 t/s |
| 24 | Qwen3.7 Plus Alibaba | 71.2 | 54 t/s |
| 25 | GPT-5.6 Sol OpenAI | 70.9 | 72 t/s |
| 26 | Grok 4 Heavy xAI · preview | 69.6 | 50 t/s |
| 27 | Claude Sonnet 4.6 Anthropic | 69.5 | 50 t/s |
| 28 | Qwen3.6 Plus Alibaba | 69.2 | 55 t/s |
| 29 | Qwen3.5 397B-A17B Alibaba · Terbuka | 68.3 | 71 t/s |
| 30 | Qwen3.8 Max Alibaba | 68.1 | 55 t/s |
| 31 | Muse Spark 1.1 Meta | 67.5 | 208 t/s |
| 32 | GPT-5.6 Terra OpenAI | 67.4 | 127 t/s |
| 33 | Gemini 3 Flash Google | 67.0 | 120 t/s |
| 34 | Nemotron 3 Ultra 550B NVIDIA · Terbuka | 66.8 | 142 t/s |
| 35 | Grok 4.5 xAI | 66.7 | 59 t/s |
| 36 | Gemini 3.5 Flash Google | 66.6 | 267 t/s |
| 37 | GPT-5.4 OpenAI | 66.3 | 60 t/s |
| 38 | GPT-5.3 Codex OpenAI | 65.7 | 118 t/s |
| 39 | Gemini 3.6 Flash Google | 65.6 | 233 t/s |
| 40 | MiniMax M3 MiniMax | 65.0 | 93 t/s |
| 41 | GPT-5.6 Luna OpenAI | 64.8 | 175 t/s |
| 42 | GPT-5.2 OpenAI | 63.5 | 90 t/s |
| 43 | Kimi K2.6 Moonshot AI · Terbuka | 63.5 | 40 t/s |
| 44 | Kimi K2.7 Code Moonshot AI · Terbuka | 61.2 | 39 t/s |
| 45 | Hunyuan Hy3 Tencent | 61.0 | 69 t/s |
| 46 | MiMo V2.5 Pro Xiaomi · Terbuka | 60.2 | 65 t/s |
| 47 | Inkling Thinking Machines | 58.5 | 81 t/s |
| 48 | Claude Sonnet 5 Anthropic | 48.9 | 96 t/s |
Podium Score adalah campuran berbobot dari lima papan peringkat. Selengkapnya di metodologi.