ArcyDev

Building things, bre̷a̷k̷i̷n̷g things, fixing them.

[ArcyDev @ArcyDev]

@IT_unhinged What a great story! I felt like I was right there with you 😂

The whole "don't touch it, nobody knows what it does" situation was hilarious. And the Brian part absolutely killed me.

You just earned a follow!

[ArcyDev @ArcyDev]

@robj3d3 0 to your first 100
followers in a week - that doesn't sound very believable

[ArcyDev @ArcyDev]

@xenappblog I really didn't need to get this kind of inspiration from X 😂

[ArcyDev @ArcyDev]

@Kalshi BREAKING: Chinese rivals will offer 3x that.

[ArcyDev @ArcyDev]

If the very first AI model had been as good as Terra 6.1, people would've lost their minds over it.

[ArcyDev @ArcyDev]

@OpenAIDevs @coderabbitai Real project use case

[ArcyDev @ArcyDev]

I compared GPT-6.1 Terra and Sol on real bugs from my game project.

Terra max found 3 additional bugs that were still in the code, while finishing faster and at a lower estimated cost.

Full methodology, results, and limitations in the article. https://t.co/htrku1ss0W

[ArcyDev @ArcyDev]

@paulette94 @elonmusk How did your parents, grandparents, and great-grandparents manage? They had far less than we do today. Excuses.

[ArcyDev @ArcyDev]

@hyphhytv @elonmusk I don't understand how anyone can reduce such a complex issue to just money.

[ArcyDev @ArcyDev]

@leonxwalker @elonmusk I think addressing this would require better support for parents, whether that's guaranteed access to daycare or preschool, meaningful financial support, or substantial tax breaks.

[ArcyDev @ArcyDev]

@jdud6611 @elonmusk It's not just about money. It's a much more complex issue. People also prioritize a comfortable lifestyle. My grandparents were much poorer than we are today, yet they had far more children.

[ArcyDev @ArcyDev]

@elonmusk Poland, nooooo

[ArcyDev @ArcyDev]

Actually, Astra 6.1 will drop on Day 28.

[ArcyDev @ArcyDev]

Day 28: @tibo drops Terra 6.1 (because it's going to be heavily hyped)

Mark my words.

[ArcyDev @ArcyDev]

I compared GPT-6.1 Terra and Sol on real bugs from my game project.

Terra max found 3 additional bugs that were still in the code, while finishing faster and at a lower estimated cost.

Full methodology, results, and limitations in the article. https://t.co/htrku1ss0W

[ArcyDev @ArcyDev]

I don't get all the hate for Terra 6.1.

Algorithm? Hello? Get it together.

[ArcyDev @ArcyDev]

Imagine this: @thsottiaux announces 👉Terra 6.1👈 AND throws in a limit ⭐️reset⭐️ on top of it.

That would be glorious.

[ArcyDev @ArcyDev]

@thsottiaux No reset, and I'm happy about it. Thanks, @tibo!

[ArcyDev @ArcyDev]

@thsottiaux Maybe people will finally stop crying about wanting GPT-4o back.

[ArcyDev @ArcyDev]

@robj3d3 @AnthropicAI 1 prompt with Fable: 70% of your limit gone. Best regards.

1 prompt with Terra: You've just used 1% of your 5-hour limit.

[ArcyDev @ArcyDev]

GPT-6.1 Terra just launched. Here’s what it’s built for:

1. Interpreting images, charts and technical drawings.
2. Working across apps: reading invoices, filling forms and entering data.
3. Drafting and editing proposals, emails and reports in clear, natural language.

Source: OpenAI

[ArcyDev @ArcyDev]

People see numbers like these and still call it boring‼️

👉 Some of you are just hating for the sake of hating and farming engagement.

You post benchmark charts and ask, "Why would I even need this model?"

Seriously? A model that's roughly 🚨 9× cheaper per task still isn't enough for you?

P.S. I added the emojis myself, not AI 🙊

[ArcyDev @ArcyDev]

⭐️ GPT-6.1 Terra’s most interesting number: roughly 9× lower cost per task than Opus 5.5 in OpenAI’s Agents’ Last Exam tests.

The benchmark covers professional computer work, including turning architectural drawings into a 3D model and creating a time-coded shot log from a film.

🚨 Standard API pricing per 1M tokens, input / output:

• Terra 6.1: $1 / $5
• Sol 6.1: $2 / $10
• Opus 5.5: $4 / $20

That makes Terra’s tokens 50% cheaper than Sol’s and 75% cheaper than Opus’s. The larger benchmark saving also comes from using fewer tokens.

For repetitive agent work, that’s enough of a gap to make me test Terra on a complete workflow: result quality, retries and total cost.

Source: OpenAI