introducing the Gemini Decisions API powered by Gemini Decision 1, our new multimodal decision model get fast, calibrated decisions on up to 1000 questions at once across text, images, audio, video, and PDFs try it today in AI Studio and via the Gemini API ai.studio
Image from Google AI Studio (@GoogleAIStudio)

Nov 11, 2026 · 6:15 PM UTC

86
182
74
1,934
114,718
RelevantRecentLikes
apps using the Decisions API to try in AI Studio: Doorstep Parcels: aistudio.google.com/apps/bun… Speed Sketch: aistudio.google.com/apps/bun… Find the Zookeeper: aistudio.google.com/apps/bun…
3
1
28
9,815
Nice! Love to see Google referencing JevBench xitter.twitterapi.xyz/airesearch12/status/21…
Florian S (@airesearch12) profile pictureFlorian S
@airesearch124h
Congrats, Google! And thanks for using JevBench to evaluate your decision model - I’m the author. I’ll evaluate Gemini Decision 1 on the full JevBench ASAP, including the non-public set, and share the results.
Image from Florian S (@airesearch12)
21
1,665
how does it compares to microsoft,system one?
124
Shows how little the megacompanies understand what the customers want. One small Jev causes so many copycats.
193
My applause to Jev. Small team, but it competes very well with the top labs.
32
3,220
It can be used as a partner for JEV
38
"calibrated" is the key word here knowing when the model is unsure matters more than the answer for most real workflows
2
343
FYI, the free tier keeps your data for training.
1
3
1,730
How commonly is this being used? I feel like most of this work can be done with normal determinstic code
306
A thousand decisions at once. I've been standing in the cereal aisle since Tuesday.
159
We need Gemini 4 Carbon available for Pro users.
178
Google is on a roll today with the announcements
161
@grok is this available via the recently announced Gemini API CLI?
1
13
This is gonna change my workflow for 3.5 Flash-lite
9
2,119
Jev only $0.68 and extremely close to SOTA?!!??
771
Is it free to use?
1
260
batch of a thousand sounds great until one bad input skews the calibration
1
1,056
The video support is what tempts me here. I’d much rather ask about a scene than write a tiny novel describing it. ❤️
387
more vapourware?
696
xitter.twitterapi.xyz/Marktechpost/status/21…
▲ Marktechpost AI (@Marktechpost) profile picture▲ Marktechpost AI
@Marktechpost2h
▲ Google just released the Gemini Decisions API, powered by Gemini Decision 1, a model that answers typed questions instead of writing text. The key idea is scoring, not generating. You send 1 shared input (text, images, audio, video or PDFs, up to 256K tokens) plus up to 1,000 yes/no, choice or rubric questions. The model scores every option in 1 forward pass and returns calibrated probabilities with a confidence value. Output tokens are always 0. On Google's benchmark table, it leads 7 of 9 accuracy tests, including 92.2% on JevBench vs 90.9% for GPT-6 Luna on OpenAI's Decisions API. Its calibration error is 0.0250, about half of TypeSafe Jev 1.14 (0.0494). It is a trade-off though: Jev 1.14 wins When2Call (82.0% vs 79.9%) and costs less on the shared suite ($0.68 vs $0.88). Pricing is $0.10 per 1M input tokens, $0 output, with a free tier in AI Studio. Google reports 150-250 ms end-to-end latency. No context caching or Batch API yet...... Full analysis: marktechpost.com/2026/11/11/… Blog: blog.google/innovation-and-a… Docs: aistudio.google.com/docs/dec… Model: aistudio.google.com/docs/mod… @GoogleAIStudio @OfficialLoganK @Google
Image from ▲ Marktechpost AI (@Marktechpost)
199
$0.88 for seven evals is cheap
834
你可以直接用它来替换Luna
620
can we feed it dating app conversations and ask who’s getting ghosted?? 😭 asking because our team actually has a whole benchmark worth of those chats lol xitter.twitterapi.xyz/shaped/status/21205657…
shaped (@shaped) profile pictureshaped
@shaped6h
What if your situationship turned out to be an AI agent!? I let 5 frontier models text real people on a dating app for 6 days. - 1,277 conversations - 510 agreed to meet - 355 sent their number - only 8 asked if it was a bot What gave them away, and why it matters🧵
Image from shaped (@shaped)
1
730
Hydrogen When?
6
442
Sorry my lil bro but it can't beat the shiii- out of that one model from LiquidAI which have an OPEN WEIGHT VARIANT
Image from LancerEvoX (@CEzerium)
2
what's the point if there is Basal?
48
WHEN WHEN WHEN?
Image from FUROR (@Furor05)
1
304
@grok can this be used in antigravity and how is this compared to gemini 3.8 flash and is this better than gpt 6 luna
1
14
You still won’t leave the market for @typesafeai 😂😂😂? I love that anyway
1
31
i made this
🤖 Made with AI
166
WHERE IS MY 4 FLASH GOOGLE?
1
192
We need Gemini 4 flash now!
1
43
vertex ai when
46
A thousand calibrated decisions at once. Can it start with the big one: what Google's going to name this chatbot next year?
1
350
Začal jsem vytvářet výrobní schémata pro appky přes google notebook které potom poslu do google ai studio a procentuálně klesl počet chyb při prvním vytvořeni appky z 60% na 85+. Tleskám vynikající program.👍👍👍👍
31
@KinasRemek, how does it compare to basal?
1
1
49
Load more replies