AA Intelligence Index: DeepSeek V4.1 Flash at 45, 3 points behind GPT-6.1 Terra
On the Artificial Analysis Intelligence Index, Terra Max scores 48, MiMo 46, and DeepSeek V4.1 Flash 1017 Max and GLM-5.3 both 45, while DeepSeek's peak output price of $1.20 is about a quarter of Terra's.
So if you dig into GPT-6.1 Terra, it's not just that OpenAI made inference cheaper. It's how little performance separates it from the best open-weight models. 👀 Artificial Analysis Intelligence Index shows ... 🔒 GPT-6.1 Terra Max → 48 🔓 MiMo-V2.6-Pro → 46 🔓 DeepSeek V4.1 Flash 1017 Max → 45 🔓 GLM-5.3 Max → 45 But if you also compare API output prices... 🔒 Terra → $5.00 / million 🔓 GLM-5.3 → $4.40 🔓 DeepSeek → $1.20 peak / $0.60 off-peak 🔓 MiMo-V2.6-Pro → $0.87 So OpenAI has pushed a strong proprietary model down to 📥 $1 / million input 📤 $5 / million output But it still hasn't created a huge intelligence gap over models whose weights you can download. DeepSeek is only 3 AA points behind Terra, while its first-party API costs 📥 $0.30 / million input peak 📤 $1.20 / million output peak And half that off-peak. It's MIT-licensed too. That's the bigger fight. Open-weight AI has become good enough that closed labs now have to compete on both 💰 price & 🧠 performance