GPT-6.1 Terra: delivering fast and capable computer use at a lower pr…

@hii_ajseven Reply ♥ 0

@OpenAI model is banger !! I am learning more about AI, how context works like that lets connect.

Original

@KennnGe Reply ♥ 0

@OpenAI 'capable model for your whole workday' ok but can it sit through my 3pm and nod at the right moments

Original

@ryderjj89 Reply ♥ 0

@OpenAI Yeah sorry Terra is terrible. One simple task to update one app took all of my 5 hour limit and another 5 hour limit that I bought for the $5 and it's still not done. That's just unacceptable. I don't care if it's the $20 plan, it shouldn't go through usage this fast.

Original

@KajetanMastela Reply ♥ 0

@OpenAI Nice performance https://x.com/KajetanMastela/status/2114908408176402568?s=20

Original

@NoxidTng Reply ♥ 0

@OpenAI Ayyy @claudeai look at this clown

Original

@krsuncom Reply ♥ 1

API로 돌려볼 분들은 가격표 각주 하나 꼭 보세요. 기본은 입력 100만 토큰당 1달러, 출력 5달러인데, 한 요청 입력이 272K 토큰을 넘으면 그 요청 전체가 입력 2배·출력 1.5배로 계산돼요. 컴퓨터 사용은 스크린샷이 턴마다 쌓여서 컨텍스트가 생각보다 빨리 불어나니까, 오래된 화면 캡처는 잘라내고 272K 아래로 유지하는 것만으로 비용이 꽤 갈려요. 급하지 않은 대량 작업은 Batch/Flex로 돌리면 50% 싸고, 반대로 Fast 모드는 2배라는 것도 같이 계산해두면 좋아요.

Original

@tang_yuanl6568 Reply ♥ 0

@OpenAI 这个模型有点烂

Original

@Kamaleshsri312S Reply ♥ 0

@OpenAI Sounds cool

Original

@112263Faye Reply ♥ 0

@OpenAI Please let 4o stay, keep the access to 4o open for us. #Keep4o #keep4oAPI #Love4o #Save4o #Keep4o4ever

Original

@D_D_12741 Reply ♥ 0

@OpenAI All I care about is when the next Astra drops.

Original

@nknow668 Reply ♥ 0

@OpenAI So good

Original

@Noobwork Reply ♥ 1

@OpenAI Computer use is the part that actually changes how I work. Chat is nice. An agent that can see the screen and click through the boring apps is the unlock. Glad this just got faster and cheaper.

Original

@Aidan_Create Reply ♥ 0

@OpenAI The thread is a gold mine

Original

@iambasitshah Reply ♥ 0

@OpenAI Faster and cheaper computer use changes the economics of automation for small teams. Curious how it handles long tasks with interruptions and permission boundaries.

Original

@kazzzz520 Reply ♥ 0

@OpenAI You guys should fix this first.

Original

@phatggg Reply ♥ 0

@OpenAI you guys need to make sol better, not this lol

Original

@AdamJWenborn Reply ♥ 0

@OpenAI Keen to see what an agent costs to run all day on Terra.

Original

@wuwenzhousear Reply ♥ 0

@OpenAI AGI agin？

Original

@wuwenzhousear Reply ♥ 0

@OpenAI why just flash or pro？

Original

@ildoraStudio Reply ♥ 0

@OpenAI It's useless release. If I need to really use GPT model then GPT 6.1 Sol is much better option than 6.1 Terra. The disappointment is that the intelligence gap between Sol & Terra is more than the Astra & Sol. 6.1 Sol is still better value. I would prefer Haiku 5.5 over Luna 6.

Original

@gui_song65431 Reply ♥ 0

@OpenAI Matching Astra at 1/10 the price is the headline, but computer use lives or dies on recovery. What's the success rate when a pop-up or login wall breaks the flow mid-task? That number matters more than any benchmark.

Original

@APIMart_ Reply ♥ 0

@OpenAI Curious how Terra handles multi-step tasks when something fails. Does it recover automatically or ask the user to step in?

Original

@cosmgasm Reply ♥ 0

@OpenAI Computer use is the real unlock. The day an agent fills out my end-of-shift report on ship Wi-Fi without timing out, I'm framing the screenshot.

Original