@ArtificialAnlys That 4 point gap to GPT-6.1 Sol is the whole pitch. OpenAI says Terra finishes tasks over 3x faster than Sol, and for an agent clicking through apps all day, speed beats 4 index points.
OryginałOpenAI wydało GPT-6.1 Terra, przesuwając granice efektywności kosztowej. To pierwszy model klasy Terra od czasu GPT-5.6, który jest inteligentniejszy za mniej niż jedną trzecią ceny. Cena za token wynosi połowę lub mniej w stosunku do GPT-5.6 Terra – 1$/5$ za milion tokenów wejściowych/wyjściowych w porównaniu do 2$/12$ w poprzednim modelu. Zapis do pamięci podręcznej również spadł o połowę, obecnie wynosi 1,25$ za milion tokenów. Zbuforowane dane wejściowe to jedna czwarta, czyli 0,05$ w porównaniu do 0,20$ za milion tokenów. Najważniejsze wnioski: ➤ Optymalność Pareto w zakresie inteligencji i kosztów
0
0
0
0
Komentarze 25
-
-
@ArtificialAnlys im just going to keep using 6.1 sol bro this makes 0 sense
Oryginał -
@ArtificialAnlys The 50 point hallucination drop to 38% is a substantial shift for GPT-6.1 Terra.
Oryginał -
@ArtificialAnlys Seeing Terra's efficiency makes me reconsider budgetary allocations in a way that feels almost philosophical.
Oryginał -
@ArtificialAnlys where does it land on the coding agent index ?
Oryginał -
@ArtificialAnlys Better than grok 4.7 lol
Oryginał -
@ArtificialAnlys
Oryginał -
@ArtificialAnlys Not gonna lie, Seems like a pathetic launch based on these numbers
Oryginał -
@ArtificialAnlys cost frontier is one axis. on the people's side, openai's best liked model right now is gpt 6.1 sol at #16 of 317 (6.8 vibe from 15.8k voices), while gpt 5.6 terra sits at #74 with a 6.1. will be watching whether 6.1 terra closes that gap with actual users
Oryginał -
@ArtificialAnlys Per Artificial Analysis, output pricing falls to $5 per million tokens from $12, and the cost to complete its benchmark tasks at max effort drops to $0.40 from $1.40. Cached input is cut to $0.05 per million tokens from $0.20.
Oryginał -
@ArtificialAnlys Good chart, thanks
Oryginał -
@ArtificialAnlys For computer use, tokens per finished task matters more than price per token. A model that needs fewer screenshots and retries can still win on the real bill. Would love to see cost per completed task next to per-token pricing.
Oryginał -
@ArtificialAnlys cached input going from $0.20 to $0.05 is the line i'd look at first. agent loops resend the same long context every step, so on long runs the cache price probably moves the bill more than the $1/$5 headline.
Oryginał -
@ArtificialAnlys bruh?
Oryginał -
seeing the exact same ordering in codex on farmerbench (three 6-hour seeds each at medium effort): gpt-6.1 sol: 74.1 mean at $0.47 a run (159 points/$) gpt-6 astra: 66.6 mean at $1.79 a run (37 points/$) gpt-6.1 terra: 62.2 mean at $0.22 a run (286 points/$) gpt-6 luna: 15.7 mean at $0.08 a run (189 points/$) at 286 points per dollar and 2m 56s a run, terra sits right on the pareto frontier between deepseek v4.1 flash ($0.13/run, 46.8 mean) and 6.1 sol ($0.47/run, 74.1 mean).
Oryginał -
@ArtificialAnlys This is the same Sonnet fumble. Dinner and more expensive than Sol. No place for it.
Oryginał -
@ArtificialAnlys @grok if this has 20 point jump in Terminal-Bench 4.0 eval and matches gpt-6.1 sol at max effort, does that mean it is excellent for coding work given cost?
Oryginał -
@ArtificialAnlys Benchmarks look great. I'll give Terra, Claude and Grok the same real task today and post cost per task.
Oryginał -
@ArtificialAnlys umm https://x.com/garyfung/status/2114849687396286696
Oryginał -
@ArtificialAnlys Opus 5.5 still seems like the best
Oryginał -
@ArtificialAnlys Cool, OpenAI's really been pushing on cost efficiency these days. Looking forward to seeing its actual performance.
Oryginał -
@ArtificialAnlys 更强还不到三分之一成本，这波确实狠
Oryginał -
@ArtificialAnlys Not the most impressive upgrade, more like a tiny improvement. Competes with 6.1 Sol depending on the reasoning level. xHigh seems to be particularly lazy
Oryginał -
@ArtificialAnlys Virei ateu de benchmark
Oryginał -
@ArtificialAnlys who is terra for? it's 6 points under sol on the coding agent index, and for agent work it still looks underwhelming next to sonnet 5.5. is the case just low effort at volume, where it sits on the cost frontier?
Oryginał
스크롤하면 더 불러옵니다