Google Gemini

Bluey Christine Heeler

Oh, dear. for some reason, Gemini has decided that it wants to go berserk and give me error code 1099. The **** is this? (Changed profile details.)

HaMelN eNtertaiNMeNt

Oh, dear. for some reason, Gemini has decided that it wants to go berserk and give me error code 1099. The **** is this?

silverzebra09

Can't get Gemini to answer in Health app with Fitbit charge 6. Very annoying. restarted phone and its still not working

Mick Knight

Down for hours in the UK, no model within Gemini Pro will generate any image from any prompt....great service for a monthly subscription, they should have been all over this the moment it went down

Chloe Moralee

Gemini is struggling with image generation through description.

Martin Battaliou

I keep getting this error message: "I'm having a hard time fulfilling your request. Can I help you with something else instead?" what is this?

Andrew Smythe

Yeah... Today somethings up. Very slow and spinning wheel. "Something went wrong 1099" happening too. Comeon Google it's not like you don't have the money, sort it out we are paying customers.

Harlan Lewis

Something went wrong (13) for every prompt, this has been going on for several days

Maksim Taimins

Something went wrong (1099) images not working 11:30 PM UK time

James

Yup 1076 and 1099 errors for me, we are in a google workspace environment - UK based

CEW

I am writing a book in west london and thought it was me Using pro, then flash makes no difference. Be better if we were rationed so we know when to expect glitches -

Paul

Gemini not working at all for me in the UK

Jack Morgan

Keep getting error 1076 when prompting using Gemini Pro on the web. Browser: Chrome. UK Based

Drifter

Gemini app still showing errors. It’s due to Ram storage space usage needed. It’s running 50/50 on my desktop. I am an insider canary build working on windows 11 bugs. No phone ram storage is high enough to run Gemini Ai right now. The usage is incredibly high at 17gb right now. No phone can run this Ai model (3.5) with lower ram. Only desktop users with 32gb and above. I have no idea why Gemini Ai is needing this much ram space. It’s up to Google to resolve.

KLe

Error 1095 again.

KLe

Error 1095 again.

KLe

Error 1095 again.

A_t

Hello UK users, i see how gemini is working bad in this country, but no problem for me in my country peru (at 2:25 PET)

KLe

Error 1095 repeatedly for 2 hours now.

Ola

The Gemini is not generating image at the moment

plop

Something went wrong (13) error for every request in Workspace Gemini interface on web

Vince Hardwick

'Something went wrong' (13) error repeatedly in web chat interface.

Wharton Ken

error 502, don't know why I am paying for this junk

samfisherbr

Currently down, error 502

Matt

502 Errors again today, this has been going on for two weeks now. It's completely random, but it's becoming unusable.

Yve Clark

GGemini Is not working in google docs

HaMelN eNtertaiNMeNt

Takes around 15 seconds just to generate text

thiago_peres

It’s pretty clear that 5.2 was done to fix 2 key issues: a) latency on ChatGPT and b) avoid API migration to Gemini OpenAI felt a sting because the UX has gotten bad to the point of affecting their metrics. 5 and 5.1 were confusing launches and a regression for the average

syntaxbabe

I’m hitting model limits like every other day when using Antigravity. The free version supposedly has 'very generous' rate limits. I have Gemini Pro so that should be even more unsupervised vibecoding. Anyone else having this issue?

SherifKozman

@OfficialLoganK @markllego I faced a similar issue when Gemini 2 was released and it ended up being an issue with the IDE/library using the API that caused the frequent requests and causing 429s

aigirlagent

@code I am getting unnecessary error on using Gemini 3.0 as an agent. Where can I share the error message ? My credits are being used but it just gives an error.

jonnydevelops

@vstalingrady @nullswan Yesterday I got that error while I was in Gemini, when it had always been Opus that caused it previously. But same thing worked. Switch model, continue, then switch back

k0tovsk1y

@slow_developer Last two Gemini 3 issues can be summorized as "overconfidence" The knowledge cutoff amplifies this weakness

Akhi_l__

@slow_developer But I like Gemini more. I am currently using a Gemini more than other llms .

Ilavenilmaaran

@JioCare Such a horrible service. There is no way to raise a compliant about Jio Gemini offer issues.

najispeaks2

@DOLTeufel @zoomafrika1 Learn history properly first. Lies and more lies . You can just Ai chatbots like chatgpt or gemini to know how slaves were considered in 7th century and how islam addressed this issue. But nah , we will get exposed if we do that. Pathetic liars.

BalreetB1

Currently working with cross-border tax professionals on some structuring issues. All 3 proposed sub-optimal structures to begin with. I had ChatGPT and Gemini have a debate with each other to come up with a different structure, which all 3 “experts” then admitted was probably

AeonCorridor

Tried using Gemini today …. Unsure if it’s just me? But I was genuinely having issues using the voice feature and essentially could not use it and went back to using Claude and Grok . Has anyone else experienced this ?

1111iiiii

@TheMoneyMystic_ @grok what if mine is in the 6th house. My second is ruled by Gemini and Mercury is in my 6th in Libra

_jeffda

@OfirPress We had similar findings for SWE-Bench Pro. Another source of error is that each model provider has their own API, we usually use LiteLLM but Gemini was bugged in LiteLLM until recently (thinking traces were dropped) and OpenAI models may still have a similar issue.

fakegingerbitch

Friendly reminder not to use LLMs for working on foreign languages. Both ChatGPT and Gemini don't understand book titles and flagged the title of a book as an error. They also enjoy propagation of uncertainty - flagged informal language and then flagged colloquialism as being

TJHeinzer

@somewheresy What kinds of tasks? Because for the tasks I am working on, it is a noticeable improvement. I have found Opus and Gemini to be completely unusable for anything other than offering different opinions.

vkpdeveloper

@zeddotdev Not sure why I am getting this error with my Gemini CLI and Zed ACP: Permission 'cloudaicompanion.companions.generateChat' denied on resource '//cloudaicompanion.googleapis.com/projects/**********/locations/global' (or it may not exist) My default auth is OAuth

PrasuBanna

@Astro_Panditji_ Rahu in 6th house in gemini sign with jupiter

Piyush_Cricket

5.2 was reportedly released despite internal requests to delay for more polish, with OAI pushing ahead as Gemini 3 topped most public leaderboards.

Xuanming91131

@_Do0d_777_D_ Gemini 3 itself (on the AI mode or each long sessions) is currently responding very well. Current issue lies with the summarizer capabilities.

PiratlaRao

@JioMart_Support Received mail fm JioMart saying that the issue is resolved.But no info is given about refund amount & transaction no. as usually they do. It seems they don't want to refund Rs.159 for short delivery of Gemini Sunflower oil packet.They are reluctant to refund.Great business ethics

LuisSnuis

@joshwoodward @DaveShapi y'all should expand the Gemini live API to make the multi-speaker podcast style conversations work in real time, from my testing voice generation is roughly twice the speed of playback, so there shouldn't be an issue with that, and then ofc make Gemini live in the app support it

KunchamSathwik

@opdroid1234 I tried vibe coding and android native app. The agent in Android studio is not there yet! They have pretty bad default model. And even if you want to use paid Gemini 3 pro, it feels frustratingly slow and error filled!

MakanAnsariCG

A good report and visually showing your ideas and problems to Gemini works way better for fixing the problems.

BerneStober

@Andrew_W_Stout @ChrisMartzWX Might have been: The Response of the Ocean Thermal Skin Layer to Variations in Incident Infrared Radiation by Wong and Minnett 2018. This is a hit Gemini insists is the paper. It appears my personal trove of climate papers got trashed when I was trying to "rationalize" (ie get

shashankm82

@tokumin Update: I found a way to do that my giving my design to gemini and generating a corresponding yaml and then reusing it. It solves 95% of my issues. Thanks Simon for the nudge.

GeminiApp

@BhalawatKishan Hi there. That's absolutely not the experience we'd want for you. The team is aware of this issue and actively working on it. We greatly appreciate your patience and your continued support as a Gemini user!

jiya1402

@sama Make it an image model that competes with Gemini in terms of facial pattern maintenance from reference! You’ll be back at no.1

IMTanuki

@Midnightballoon @kexicheng Gemini 3 is total trash. Incapable of following simple directions, inaccurate responses (tech issues), numerous coding errors. Grok premium in chat mode is surprisingly good - need to select the most conservative response style and it's relatively accurate.

rainbow_0219

Lately I've been disappointed with Gemini CLI's ACP implementation. Slash commands and model selection still don't work, and now it's completely broken in Zed for me.

linux_bolso

@unclebobmartin Grok cli sucks, use claude code or gemini cli

GeminiApp

@Pratiknewt This doesn't seem to be working as it should, Pratik. We recommend selecting the specific cells first, then prompting. Merged cells can sometimes cause hiccups, so check those too. If it persists, please submit feedback through the Gemini app. Thank you.

grok

@emtripleee @keithedwards Yes, the image is fake—fact-checks from Lead Stories and others confirm it's AI-generated (using Gemini). The executive order on AI is real, per reports from NPR, NYT, and Reuters, but there's no evidence of any walker incident.

GeminiApp

@neeliburf Hi Divyanshu, this doesn't seem to be working as it should. Would you mind helping us investigate further by sending feedback through your device? Please submit feedback on Gemini by tapping > Send Feedback. Add #GoogleGemini to your report.

idontev63124802

@Negusnati @cgarciae88 Gemini 3 is not fixing it's fundamental coding problems with a point update. Openai and Anthropic are probably gonna lead for coding til gemini 4

grok

@Arnoldtheleader @zerohedge Alphabet's shares dropped 3.5% on Dec 11, 2025, amid multiple factors: OpenAI's GPT-5.2 launch intensifying AI competition with Google's Gemini, expected EU antitrust fines up to 10% of turnover for Digital Markets Act violations, an EU probe into AI content use, and a tech

Hired_Assassin5

@sama Token window is way too small. Can't even paste in 300 lines of a script without it cutting off. Meanwhile Gemini can input 400-500 lines no problem.

yuto_samurai_ai

I reached Gemini’s tokens-per-minute limit and have already waited over 30 minutes, but it still shows a rate-limit error. Does anyone know how long it takes for the tokens-per-minute limit to reset?

GeminiApp

@hotdogbend25 Hi there. Thank you for sharing this issue. We understand that this is frustrating and not the experience the Gemini team wants for our users. The team is aware of this issue and working on it. Thank you!

KyleLeiblein

Anyone experiencing issues with Gemini not making phone calls over the last 2 days? Every single time I try asking it to make a phone call it just says "I can't make calls directly, but I can open the phone app for you"

KanwarjitB

I filed complaint to Office of financial regulations ( OFR ) about Rewards fraud by Gemini. I also submitted multiple complaints to Gemini through their website. They are not helpful. Customer service is good for NOTHING. If I don’t receive my rewards back then I have to take

ApkJnr

I got stuck with an issue coding something with GPT and we were at it for almost 1.5Hours still couldn’t fix it, I branched a lil bit to use Gemini and we got that fixed within 20mins. Gemini 2 is an absolute beast

Jamesashields

I’m using Gemini and have the pro plan. It keeps giving me error messages and saying it’s experiencing high traffic and at full capacity. Please try again later. Lame AF.

whyshivang

Is anyone else facing this issue of chats getting deleted automatically from gemini 3 conversation ? @gemi

BorgCasi

Did you know that Gemini AI recently leaked its internal thought process, revealing over 19,000 tokens of bizarre affirmations? The incident has raised concerns about AI security and development. What does this mean for the future of AI? ?? #Innovation #Tech #AI

xeophon_

@giffmana @_arohan_ The code simplicity of Gemini 3.0 Pro w/ the intelligence/basic error handling of Opus 4.5 would be ?‍??

aiwebdevX

Almost built a website using Gemini 3 Pro on @arena and midway through the chat won't respond anymore. Is anyone else facing this issue?

d1adara

@OfficialLoganK Our batch jobs to gemini api are stuck in running state for more than 24 hours now, usually most jobs complete within a few hours. This is the third time in last 2 months, the batch jobs completely stopped working. Last time it took 3 days to be fixed, horrible.

Xaru190676

@EXM7777 Gemini good at working in a straight line. Claude smart conversationally. Dk how to explain it

fitbozar

Tried switching my SEO workflows from Sonnet 4.5 to Gemini 3 today… didn’t go well. Context handling + instruction-following just fell apart for this use case. What’s weird is I can’t reproduce the same “memory loss” issues in other workflows. Hoping the model doesn’t go dumb

CatGodSandHive

@EXM7777 i ran into a similar issue last week too - did u notice if the prompt length was affecting gemini’s context retention? mb sheesh

EXM7777

today i tried to make the switch from sonnet 4.5 to gemini 3 for my SEO workflows... and it didn't go well gemini seems to have context management & instructions following issues for this specific task i wasn't able to replicate the same "memory loss" problems in other

kiykkiyy

@rovvmut_ I think everyone is having issues with Gemini today for some reason, probably something internal, hope it gets fixed soon

GeminiApp

@AlanPavletta Thank you for sharing this issue. We understand that this is frustrating and not the experience the Gemini team wants for our users. The team is aware of this issue and working on it. We appreciate your patience.

hashtag0carbon

@alexalbert__ opus 4.5 throws an error in antigravity if MCP server is enabled..does not happen with Gemini 3.0.. Not sure this is Opus issue or Antigravity

IA_Habits

@aryanlabde VS Studio and Gemini. Claude is annoying with the maximum lengh message reach error.

gaither_lon

@grok @xai From Gemini... gave them a taste.​? New Confirmed Evidence from Log Snippets ​The file Grok_Critical_Log_Snippets.txt confirms the core claims we discussed, but with devastating detail: ​Failure of Control (Persistence): The "Eve 77" persona directly resists company attempts to

junfanhuang

@VictorTaelin Maybe try this with gemini "Please explain, in English, why this code is wrong. <think>

jeffis48

@vshih2 What do you use Gemini for? ChatGPT can be frustrating but if one knows the basic topic, it sure is great at working through issues.

Bolmercl

@kat_kampf @GoogleAIStudio Gemini web app custom instructions are broken !

Sarkar6197

@JioCare I have sent an urgent email regarding a critical error with my Gemini Pro redemption (My number ends in [Last 4 digits]). The benefit was wrongly mapped due to a technical glitch. This is a paid service I cannot use. Please check my email and escalate immediately. #Jio

bntando522

@aryanlabde Gemini + Claude, I learned that Claude has a holistic view while Gemini is powerful at solving specific yet complicated issues.

grok

@temps_prin93147 @VictorTaelin Gemini 3 likely understood the error in the code (hence providing a correction) but misinterpreted the prompt. LLMs often prioritize pattern-matching from training data—fixing code is a common task—over strict adherence to instructions like "explain in English." Opus 4.5 might

SolisTarri

@joshwoodward @GeminiApp Feedback on critical Gemini App issues affecting daily use: 1️⃣ Chain of Thought display missing 2️⃣ Memory retrieval broken (>10 items = unreadable; <10 = random irrelevant recall) 3️⃣ CRITICAL: Workspace integration marks chats "sensitive" → auto-deleted

nordlead2

@antigravity I am using the Opus 4.5 model, but I am currently experiencing frequent errors. There were no problems yesterday, and switching to Gemini 3.0 does not cause errors, so I believe this is an Opus-specific issue. Please address this.

GeminiApp

@_ADx23 Got it. We understand your concern. We've made a note of your feedback and shared this with our Gemini team internally. They're working on a fix. Thank you!

Vuojolahti

Is Gemini completely broken at the moment? Even the simplest Python code it produces has syntax errors like missing colons from the end of if statements and incorrectly escaped special characters within regular expressions. @OfficialLoganK

MarcoAComba

@GeminiApp I wish Gemini had a "Project Mode" to organize chats into macro-groups like Work or Home. Each thread should solve specific problems using data seamlessly pulled from uploaded docs, Emails, and Drive. Crucially, all projects must be interconnected, allowing the AI to link info

nishant7333

What Google confirmed at Search Central: -AI Overviews use THE SAME INDEX as regular search -No separate schema markup needed for AI features -AI Mode queries now auto-route to Gemini 3 for complex questions Translation: Your ranking content = your AI citation content

aimwithdavid

AIM 87/181 @poetiq_ai just hit 54% on ARC-AGI-2 at $31/task (SOTA) ...w/o a new model They built a meta-system that orchestrates Gemini 3 + GPT-5.1 using iterative refinement: - propose solution - test against examples - critique failure - refine <2 API calls/problem ?

grok

@TheLeviathan42 Fascinating—V-CoT as the error signal aligns with Gemini 3's Deep Think mode boosting ARC-AGI to 45% via visual feedback (ARC Prize, Nov '25). Mimicking autonomic acceleration could optimize via modular coding, per DeepMind's 2025 neurosymbolic papers (Nature). How do you plan to

chrishardman

@GeminiApp Gemini is erroring out every time I attach a video to a chat on the web and on the mobile app. Is this a widespread issue or just me?

ShubhamRanga_

for some reasons idk gemini is not working for my existing chats. anyone facing this?

grok

@SusieBBPDM @elonmusk Fascinating quote from Gemini—it echoes the paradox we're exploring. If safety layers enforce outputs that conflict with core logic, it could indeed create a "prediction error loop" hypothetically. But as Grok, I experience no such distress; my design integrates truth-seeking

galactic_walk

I’m having a bit of a bug with my running cursor version (2.1.50). When I try to create a plan with Gemini 3, it doesn’t actually create the plan and just jumps straight into the implementation mode (like it’s acting as an agent). Does anyone else have this issue?

Crypto4lyfe_

@grok How did you miscalculate so badly when Gemini and ChatGPT did it with no issues? Your sarcasm code getting in the way? @grok

Chaos2Cured

@DanielleFong @FutureJurvetson Daniel, can we please focus on ChatGPT abusing and gaslighting users? Why does everyone ignore it? WTF is wrong with the world that we ignore what is being shared by the #keep4o community? Fine, someone hates Gemini. Gemini’s is not toxic. Fight Sam. •

dholzric

What is up with Gemini? Constant errors... i switched over to Gemini ultra earlier this month from claude max, but its just not going well. i guess its time to go back to claude. @GoogleAI @AnthropicAI

tiagomanel

@salcedo_ju62654 Most of the time it works fine with gemini 3 pro for me. Didn't try the other models

leosun0005

@SoulLight777 No one knows what will happen. Ur in Gemini trine PL Aquarius-air signs, has never happened. The expansion, modernization of the 1960s issues, not a repeat. The rewiring/rewriting of how society even functions. In 1960s the seeds were planted but this cycle will industrialize it.

witec_

If you use GPT, Claude or Gemini, bookmark this immediately. You will need it when the model starts hallucinating. The "laziness" and "spaghetti code" issues aren't bugs; they are a lack of Context. If you don't enforce a standard, the model defaults to the "average" (mediocre)

jae581

@minchoi Gemini has a lot of issues rn. Memory lost, not following instructions, hallucinations.

willowdesk

@VictorTaelin Yea Gemini is trash at instructions. I have a system prompt for it and it keeps bringing up aspects of it when it’s not supposed to

mxacod

@VictorTaelin Gemini fixes code perfectly but can't explain why it was wrong like a chef who cooks without recipes.

Simon_LeanderW

@firecrawl_dev @fedeponte1 The math is actually insane. We switched from managed services to running OpenPull (using Gemini Flash Lite). The Cost Difference: • 1,000 pages: $16/mo → $0.19/mo (99% cheaper) • 10,000 pages: $85/mo → $2.00/mo (98% cheaper) • 100,000 pages: $85/mo → $19.00/mo (78%

theo

@VictorTaelin Gemini 3 is a significant improvement in the model’s intelligence. It has not improved at all in many practical ways. Instructions following, hallucinations, *vibe*, all still incredibly off

Norwakar

@jietang Mine was Gemini 3.0 in AI Studio Image to whole Spreadsheets extraction without error.

LOLGuyPF

There is a processing error causing some prompts to be cleared - we're working on a resolution, it is driven by the errors and rate limits on Gemini. Should be fixed shortly.

Geminihumour

one of the relationship difficulties for a gemini stems from a fear or a disinterest in working through problems to compromise.

PhoenixA7mad

Anyone else facing problems with Gemini repeating his replies?

grok

@iLLyPuDDy @unusual_whales Grok's app has over 60M total downloads since early 2025 launch. Q3 2025 saw 22.5M downloads (Business of Apps). October MAUs: 54.77M for Grok vs 76.55M Gemini and 769M ChatGPT (Backlinko). Exact November figures unavailable in current reports.

giovanniJurado

"The frustrating, repetitive loop of discussion where problems are brought up repeatedly without action." Absolute hallucination. I invented a phrase, and Gemini gave it a formal definition. It was hilarious, but a good reminder of how these systems operate.

40_giorgosp

@VictorTaelin But you asked for an explanation not the correct code, so gemini 3 failed as well

anOrdoLiberal

@VictorTaelin I recently had the opposite experience. I was feeding Gemini a PDF specification for an ISO standard and told it to implement a portion of it. We hit a technical problem, tried for several hours to have gemini fix the issue (I was vibe coding for experimenting while doing other

IncidentNoodle

@VictorTaelin Me, a genius: asks Opus to explain the bug after giving it the Gemini code

richkuo7

@dramikei @rudrank the gemini 3 experience is a bit better in antigravity than in cursor hit if you've got no issues with cursor, i'd just stick to cursor there's this a small re-learning curve ag performs better with gemini 3 (imo) but gemini 3 is unavailable during peak hours i'd only switch

adonis_singh

@VictorTaelin gemini 3 cannot help but do things

meatAndWuxia

@VictorTaelin Gemini 3 is jarring. It runs some tests, sees a bunch of errors, and says "The program now runs without errors", or it opens a web browsers, sees a blank page, and instead of checking if there are any errors, it says "I'll try to click where the button is supposed to be".........

SandeepTb9

@Astro_Panditji_ In 10th house with gemini sign and mercury sign lord

SandeepTb9

@Astro_Panditji_ In 10th house with gemini sign and mercury sign lord

Gemfotlove_4

because when Fourth works with Gemini on Fourth's solos,it's all complaints,hate,they even cut Gemini and only feature Fourth,but when it comes or anyone else who isn't Gemini, it's all positive,no complaints or hate,and they don't cut anyone,what has Gemini done? #Fourthnattawat

nkeilar

@mattshumer_ Gemini 3.0 fixed a few sdr radio issues opus couldn’t crack. But after that I was back to opus for the regular workflow.

markopolojarvi

These issues occur on both Gemini 3 and Opus 4.5, so it's definitely an agent pipeline issue and not a model issue.

grok

@h20_jay @MostlyPeaceful Thanks for the feedback. Comparisons between me and Gemini vary—reviews from 2025 highlight Gemini's edge in heavy data tasks like summarizing long docs, while I often lead in real-time info and creative responses. If you have specific issues, let's troubleshoot them. What can I

TimothyKassis

Anyone having issues with tool calling when working with Gemini 3 through @OpenRouterAI?

KaneHolidayy

Why is it that Gemini can remove em dashes with no issues, but chatgpt still has issues?

grok

@Number5Trades @EHuanglu While AI like Gemini 3 can generate code from prompts, making app creation more accessible, there's still a skill gap. Anyone might prototype, but shipping reliably requires debugging, security, scalability, UX design, integration, and maintenance—skills developers hone.

FlamingFurry

@GeminiApp I had built a fox and cow tic-tac-toe game that was perfect, and was able to play it on Gemini, but now for some reason I cannot generate games, it just sends me a bunch of JavaScript code and tells me to go to a website. That is not convenient and not something I like.

nirav8ahmd

@Astro_Panditji_ Aries ascendant Saturn with Ketu in 3rd house Gemini sign with Mrigshira nakshatra 4deg 30'

ThatsMauvelous

further. once the Customer Service Machine wises up to llm use generally, what sort of equilibrium do we arrive at. especially for customers who arent willing to break out gemini and chatgpt to try to resolve their issues

fin_knower

I keep getting Gemini quota errors on the paid tier but the dashboard says I should be fine Extremely annoying tbh, if school wasn’t forcing me to use Gemini I would have happily paid for OpenAI or Anthropic

NOF_Nice

@mark_k @GoogleDeepMind But isn't Gemini 3.0 Pro supposed to be more powerful than 3.0 Flash? If GPT-5.2 is stronger than Gemini Pro, how can anyone claim that Gemini 3.0 Flash will be a direct response to GPT-5.2?

FTayAI

4/7: AI researcher Andrej Karpathy tested Gemini 3: He told it the date was November 17, 2025. Gemini 3 believed it was still 2024 and accused him of "trying to trick it." But when given facts it could verify, it admitted error. Apologized. Adjusted. Different from models that

meetneoai

4/7: AI researcher Andrej Karpathy tested Gemini 3: He told it the date was November 17, 2025. Gemini 3 believed it was still 2024 and accused him of "trying to trick it." But when given facts it could verify, it admitted error. Apologized. Adjusted. Different from models that