GPT-6.1 Terra 與 Kimi K3.1 Pro 規格對比，DEV·TV 評為 3 比 1

Two new models hit OpenRouter this week: OpenAI's GPT-6.1 Terra and MoonshotAI's Kimi K3.1 Pro.

Before I'd even had my morning chai, DEV·TV's channel 13, MODEL FACE-OFF, had already put them in the ring.

🏆 The result: Terra wins on specs, 3–1

💡 What stands out

• Terra is half the price. $1/$5 vs $2/$10 adds up fast at scale.
• Terra edges context at 1.1M vs 1M.
• Kimi's big win is max output. 944K tokens vs 128K is a huge gap if you generate long docs, codebases or reports in one go.
• Kimi takes video input, while Terra takes files. Pick based on what you feed it.

⚖️ How MODEL FACE-OFF scores

Each row is worth 1 point. A tie scores nothing, and a row is skipped if a stat is missing. Specs and prices come straight from OpenRouter's catalog.

This measures specs, not quality. It tells you what you get for your money, not which model writes better code. That's still on you to test.

📺 What's next on the channel

• NEXT: Voxtral Mini Transcribe 3 vs Voxtral Small Transcribe 3
• LATER: Nemotron 3.5 Embed 1B vs Nemotron 3.5 Embed 8B

New models get matched against others of the same type, so you're always comparing apples to apples.

🛠️ About DEV·TV

DEV·TV is a retro TV for developers with 14 channels: GitHub, Hacker News, Hugging Face, AI Papers, CVE, New Models, Model Face-Off, Spotlight and more. It's a single HTML file with zero infrastructure.

• #5 Product of the Day on Product Hunt
• #1 Product on IndieLink
• Open source, Apache-2.0

Built something? Submit it to Spotlight right from the app, for free.

📺 Watch: https://shouvik12.github.io/devtv
⭐ Star: https://github.com/shouvik12/devtv

Would you pay 2× for Kimi's 944K output, or is Terra the smart pick? Tell me in the comments 👇