Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber

I wonder how big the Pro model is that Google is using behind the scenes to train these smaller ones.

Going on baseless speculation, the lack of accompanying pro models with these flash releases either means: 1) the model is too big to be economical, 2) google doesn't have the compute to serve the big model, 3) their big model has too many alignment issues to serve to the public.

edit: looks like benchmarks are up on https://artificialanalysis.ai/models/gemini-3-6-flash. It's solidly middle-of-pack. However, if you want to be most fair to flash, look at the intelligence vs time per task and intelligence vs outputspeed benchmarks. This is a very fast model.

edit 2: I use antigravity from time to time and in my experience, 3.5 flash is an underrated model, so long as you know what it's good for. It's very good at frontend (much better than gpt 5.5) and it's fast, so it's a great tool for iteration. I expect 3.6 to be no different.

My hunch is Google is trying to integrate a fast and relatively cheap AI across search and every other surface of their product suite. And for that objective, a model that can move faster while being accurate and cheap enough is more important to them than producing a frontier class heavyweight model.

Google somehow managed to snatch defeat from the jaws of success with their AI products.

They literally forced me and my company out of Antigravity by phasing out AI Ultra subscription without any proper product follow-up. Antigravity IDE cannot even have poweruser subscriptions now from Google Workspace an Gemini Enterprise Agent Platform cannot be attached to Antigravity IDE.

Gemini Enterprise Agent Platform has an incredibly abysmal setup process, and if I want to limit spending per-user I have to create projects per user. The fact that you cannot activate Anthropic models on it if the billing still has free credits is almost a joke.

I was a big proponent of Google and Gemini, but they left us reeling with their abrupt product decisions. Forced us to buy $200 subscriptions directly from Anthropic/OpenAI.

It's a bit disheartening to see no comparison to other models here - and I'm not sure this pushes the curve anywhere. 3.6 flash is more expensive than GLM 5.2 - but seemingly worse, although this post is really light (lite?) on details.

It seemed for a time that Google had finally gotten the ball rolling, but I'm doubting that more and more as time passes. We'll see what happens with 3.5 pro I suppose.

Pelicans for 3.6 Flash and 3.5 Flash-Lite (Cyber isn't available to me through the API yet.)

https://tools.simonwillison.net/markdown-svg-renderer#url=ht...

Pricing per million input/output tokens:

2.5 Flash: $0.3 / $2.5

3.0 Flash: $0.5 / $3

3.5 Flash: $1.5 / $9

3.6 Flash: $1.5 / $7.5

---

2.5 Flash-Lite: $0.1 / $0.4

3.1 Flash-Lite: $0.25 / $1.5

3.5 Flash-Lite: $0.3 / $2.5

Man I love Gemini models but these kind of pricing increase is just insanse. I have a little product and I have to keep increasing the price and reduce the limits because of this non-sense, and they did not even let us use the old models in near future, so I forced to update to the new model with basically no to little improvement because I don't even need that much. Google if you can read this, it okay to release new models and change the price for them, but please please don't kill the old ones like gemini-2.5-flash-lite, because that all I ever need for my little apps with only few thousands of users.

A couple tidbits:

> Beyond today’s releases, Gemini 3.5 Pro is currently testing with partners and we plan to make it broadly available as soon as it’s ready.

> We have started our most ambitious pre-training run yet, for Gemini 4, and are excited by the progress.

It is both less intelligent and more expensive than GLM-5.2, while being closed weight.

It's kind of ridiculous how good these are getting. 3.5 Flash lite is pretty comparable to Opus 4.8 (at least for the couple tests I did) while simultaneously being 6x faster and 19x cheaper.

https://fy2zp1ri90.evvl.io/

It's scary relying on Google's models.

I have a very price sensitive workload that used to run on flash 2.5 lite - it's deprecated now.

The replacement 3.1 flash lite is a lot more expensive, but now also has a sunset date.

3.5 flash lite is even more expensive.

So the price is rising and you have no choice but to keep paying more and more.

No word about updating Jules, which is still stuck on 3.1 Pro. I get that it's probably niche but I've really appreciated basically being able to give directions to Jules on my phone, then reviewing and merging a GitHub PR fifteen minutes later. It's been great for getting some progress in on a few personal projects during my commute when I can't exactly pull out my laptop.

Anyone have any good alternatives?

Wow - Google does not even bother to show benchmarks of these models compared to the frontier and Chinese labs - only against previous versions. I'm not surprised. Having worked there for years it was amazing just how inwardly looking the company is.

The mention of an "ambitious" gemini 4 pre-train signals to me that 3.5 pro is probably a lost cause.

That being said, it seems that Gemini is still the best image analysis model, so hopefully 3.6 flash builds on this even more.

Really, what's up with Gemini still not supporting connectors/MCPs/plugins/whatever-they're-called-this-month on web? It makes it a non-starter for any kind of serious use.

I have a side business selling custom fingerprint jewelry and I use gemini nano banana to clean up customer submitted fingerprint images. This was a step I used to do by hand at 10 - 15 minutes per image and nano banana is the first model that is able to do the task (it is astonishingly good at it). I can't wait to see what the next nano banana can do, hopefully its released soon.

Their naming scheme is confusing. Branding has never been Google's strong suit and their marketing copy is pretty bottom-of-the-barrel[0]. Anthropic has a pretty clear set of models but Gemini decided to rebrand their Flash as Flash Lite (and presumably the future will see a Flash Lite Mini, a Flash Lite Mini Nano and a Flash Lite Mini Nano 3B) which confuses the pricing to high hell.

This plus the Vertex, AI Studio, Gemini, Antigravity. It's honestly too confusing to use. I need to use Gemini just to decide on which platform and which model to consider.

0: Famous Kurian Tweet: "We're announcing Duet AI for Google Workspace will now be Gemini for Google Workspace. Consumers and organizations of all sizes can access Gemini across the Workspace apps they know and love. We're introducing a new offering called Gemini Business, which lets organizations use generative AI in Workspace at a lower price point than Gemini Enterprise, which replaces Duet AI for Workspace Enterprise."

Google seems to have anorexia when it comes to model intelligence. They have an internal hard constraint on price per token it seems, and they are trying to squeeze out intelligence with limited compute.

I wonder if there is something with their TPU cycles that makes them want to postpone training a new model. My guess is that they have been on the same base model for 6 months and they may have waited for the next gen TPUs to train Gemini 4, which greatly limits how much intelligence they can increase and forces them to do cost efficiency increases.

Unfortunately says more about how competitive 3.5 pro would be today at the frontier if they forgo it for 3.6 flash.

3.5-lite is the real showpiece here; agentic models of this size are a huge value-add for 90% of knowledge work agent tasks

So it's GLM-5.2 performance for almost twice the price.

That said, the speed looks really good. I think it's competitive with Fireworks's GLM 5.2 Fast, although Fireworks is still cheaper.

Google has not changed. Following two facts are like tautologies by now.

1. Their AI efforts are very fundamental research oriented. They are really good at it.

2. Their productization sucks. The end products gets little attention compared to competition. It can be canceled at any time. You should never build anything around Google only APIs, AI or not.

Google really needs to get their product strategy together. The discontinued gemini-cli and introduced antigravity-cli which is a downgrade IMO and the sooner they can partner up with AWS and release the gemini models via Bedrock the easier corporate/business which has strict data protection rules can use their models and make them available for internal engineers.

It's a one thing to research and improve the model, but if they ignore the ease of access and multi-availability of their models in different ways they are going to fall behind again.

I'm more excited for 3.5 pro. Gemini has fallen behind in some areas, but is still one of the best multimodal models.

Has anybody found any models better at image or audio analysis?

Why would 3.6 flash perform a little worse than 3.5 flash on Artificial Analysis Coding Index...

https://artificialanalysis.ai/models/gemini-3-6-flash?intell...

Wow, Laguna S 2.1 (released today) just destroys Flash-Lite underly and completely. What a weak and embarrasing release from Google.

Tons of guardrails, lazy model, super confusing plans, expensive 3.5/3.6 flash and lite and 3.5 pro MiA?

Rough patch for google ai

Pelican svg and a near-perfect 3D MacBook at max effort for $0.16, about a fifth of Fable's price.

Fable 5 still wins on detail with no visible errors, but it's close. And this isn't a memorized pelican;

https://playcode.io/blog/macbook-svg-benchmark#gemini-3-6-fl...

In other good news "the model has been trained to minimize refusals for beneficial uses.".

Otherwise, this news feels like a tiny incremental improvement on Gemini Flash series to make it more efficient with token usage, subagent and cost. Nothing big.

Regarding their benchmark scores on CyberGym, I wonder why they didn't compare their 3.5 Flash Cyber model with Fable 5. I mean they included Mythos and GPT-Cyber, so why not Fable 5 too?

They also mentioned Gemini 3.5 Pro is in testing and its about to become available very soon. Another thing maybe worth discussing is the announcement of pre-training Gemini 4. Sadly, not much technical details to discuss on. Many comments in here seem to mostly be about how Google is behind the others, but honestly, is it really worth the investment to be #1 in Artifical Analysis every week?

From my experience, this thing is crazy fast.

Spawn 10 on the same problem and have them debate to reach a consensus, you’ll get Fable-like results but 100x faster.

3.5 Flash-Lite seems available in US region, as was 3.5 Flash; but 3.6 Flash looks Global only so far when pinging. If Google employees are watching, will this issue go away?

Gemini 3.5 flash-lite is more expensive than Gemini 3.1 flash-lite. Every upgrade is getting more expensive.

It is 17% more token-efficient than 3.5 and performs significantly better in coding and tool usage benchmarks.

It is also cheaper than 3.5:

> This enhanced efficiency is also combined with a lower price than 3.5 Flash. At $1.50/1M input tokens and $7.50/1M output tokens, 3.6 Flash reduces the overall cost per agentic task, making agents more cost-effective to build and run.

Bottom line it looks about on equal footing with GLM 5.2 in terms of both overall intelligence and cost per task, while being significantly faster (in fact it is the fastest model on artificial analysis as of rn [0])

[0] https://artificialanalysis.ai/models/gemini-3-6-flash

I struggle to see any value in this when DeepSeek is still a thing.

I have just tried to switch to 3.6 instead of 3.5 in antigravity and it seems to constantly spit "critical instruction: STOP CALLING TOOLS NOW. YOU MUST WAIT FOR WAKEUP. ". I think I will switch back to 3.5

Kind of excited about this. 3.5 Flash on Antigravity has surprised me recently on a hobby project. When given opportunity to plan, it can deliver on tasks that would take me a while on my own and generates responses at blazing speeds - compared to what I'm used to at work with Opus 4.8 (granted I don't use Opus 4.8 on my hobby projects so just anecdotal). While with Gemini CLI I would just watch it run in circles and run out of 5h allowance before anything useful is produced (or even approached).

Proof-of-life release while they figure out how to have a competitive frontier model release. My hunch is they pushed too far in the "omni" model direction, that they made something so ungainly, it wasn't as good for normal tasks.

Spent half an hour just now benchmarking it against my current 3.5 Flash pipeline excited only see it regressed slightly (0.1% - 0.2% at most, for feature extraction work)

Seems like this is mostly a cost play by Google, hoping this doesn't bring 3.5 Flash capabilities to an end of life, and that 3.6 catches up or gets better.

Always happy to see new Gemini releases as IMO Antigravity Pro 16.67/mo plan (Annual) is still the best plan available and have been pretty happy with Antigravity IDE.

If it wasn't for Gemini/Antigravity I'd have to go with a Max Claude plan, as it stands now I can get by with just a Claude Pro plan to get Opus when I need it, whilst using Antigravity as my day-to-day workhorse.

Unfortunately Gemini Flash became too expensive to use as a general purpose model (i.e. for AI features in Apps), luckily there are plenty of cheaper Chinese models to fill that gap now.

I often use Gemini free web chat because it's generally quite good at web search-related questions (apparently it has direct token-level access to the Google Search index) but I noticed in the last two weeks output quality of 3.5 Flash seriously degraded. Maybe they were switching over systems.

Feels like they released this to ride the wave of press of GPT-5.6, Kimi K3, and Qwen 3.8. Doesn't feel like Google has much substance with this post except a bump in version and tweaked their pricing.

Don't know why are they even pursuing Gemini. Just download the Kimi, call it Kimini and serve it on your GPU. Maybe then train next architecture based on this!

Here's the issue:

GLM 5.2 is better, also cheaper, and almost as fast.

So essentially, a big L for Google. Combine this with them not being able to produce a frontier model this generation... hmm implications

Gemini 2.5 Flash-Lite has been my go to for cheap document processing at scale (especially with 50% off batch mode), but they are really boiling the frog with pricing increases with each version:

gemini-2.5-flash-lite: $0.10 input / $0.40 output

gemini-3.1-flash-lite: $0.25 input / $1.50 output

gemini-3.5-flash-lite: $0.30 input / $2.50 output (a 6.25x increase over 2.5!)

Now watch them deprecate Gemini 2.5 Flash-Lite in the coming months...

3.6 Flash scores exactly the same as 3.5 Flash on the Artificial Analysis index. Better on some tasks, worse on others. Mostly within what I'd consider the noise window. Looks pretty much indistinguishable from 3.5 Flash, at least on these benchmarks: https://artificialanalysis.ai/models/gemini-3-6-flash

Looks like 3.6 Flash is the first model with their newest pretraining run (cutoff date is 2026/03), long after 2.5 series.

I deeply wish Google would focus on models like Gemma. Small, powerful, open-weight models you can run on phones or regular computer hardware.

IMO Gemini has the best free tier models/app for everyday use. Muse-Spark is perhaps just slightly better, but has none of the connectivity to my GApps (for things like “create a recipe in my Google Docs from this image”).

Plus they are probably running these things on every Google search so saving tokens is a huge win for them.

For anyone wanting a faster overview: I ran the Gemini 3.6 Flash and 3.5 series release notes through NotebookLM and generated a short video summary. Link: https://www.youtube.com/watch?v=SUFBhvQ2tY4

LLM reception is truly extreme, even worse than AAA game releases.

Ever frontier lab lived it at least once : missing the frontier by a few months triggers extremly negative reactions, then you take back the lead for 2 weeks, and the hype cycle repeats.

So 3.6 Flash is a somewhat of an admission that Google miscalculated by charging 3-5x for 3.5 Flash what it did for 3.0 Flash (3x input and output costs plus large token inefficiency changes) despite only modest improvements?

3.5 Flash Lite is only a hair cheaper than 3.0 Flash, but I think 3.0 Flash is a massively more capable model?

Pretty underwhelming, as expected honestly. I don't want to know what morale is like at DeepMind right now.

It does seem like their releases are getting closer together. I get the feeling they realized they were trying to roll out to their entire ecosystem and now they’re focusing more just directly on the AI model itself. I think give it a little time and they’ll start to be one of the competitors too.

I'm a big fan of the Flash-Lite models. They're exceedingly fast and deliver great outputs for high volume use cases where you need to process requests at scale. Can't wait to try the newer version.

Flash Lite: 0.3/m and 2.5/m

Deepseek Pro: 0.435/m 0.87/m

That's wildly ambitious pricing by Google. You can maybe get away with spicy pricing at the SOTA edge but at the lower tiers everything is a lot more price sensitive.

"..and in some benchmarks like DeepSWE by Datacurve, we observe up to 65%, all at a lower cost per output token."

"3.6 Flash delivers higher precision with fewer unwanted code edits and reduced execution loops, as seen in DeepSWE (49% vs. 37%)"

So which one is it? 65% or 49%?

tl;dr: 3.6 flash is a bit smarter than 3.5 flash, but also a bit more expensive.

My results [0] put Gemini 3.6 Flash at the top.

3.6 Flash high has same $1.5 input price as 3.5 Flash, but output is cheaper from $9.0 to $7.5.

Google said 3.6 Flash is more token efficient, but in my tests it's actually LESS token efficient[1] than 3.5 Flash, so despite the output price reduction, it still costs more.

[0]: https://aibenchy.com/compare/google-gemini-3-6-flash-medium/...

[1]: https://aibenchy.com/compare/google-gemini-3-6-flash-high/go...

I was expecting 3.6 Pro. It's been so long since the last Pro model...

Are they comparing 3.6 Flash to 5.6 Luna and losing? That's ruff.

I read this as a soft let down to not expect too much from 3.5 Pro.

> We have started our most ambitious pre-training run yet, for Gemini 4, and are excited by the progress.

The benchmarks are not particularly impressive. I suppose they needed to release something since the long pause. But not clear why would I use it now.

"We made 3.6/4 Pro, but it sucks, so this is the distilled model" vibes.

I have liked using their consumer products but they don't make it easy, that's for sure.

A lot of disappointment here in the comments, but models like these aren't meant to compete with the likes of Fable or GPT 5.6.

I use 3.1 Flash Lite regularly to classify listings on eCommerce websites. It's great for this task - fast, cheap and accurate.

In fact, it was the single best model we tried in terms of the speed vs accuracy vs price tradeoffs - including the Chinese models.

Of course, 3.5 Flash was more accurate but the 5x cost increase couldn't be justified.

3.5 Flash Lite sounds like it could be a strict upgrade for our use case, without a significant increase in costs or drop in speed.

It's not GPT-6 but it's not trying to be. It's a completely different tool and great at what it does.

Glad to see the price is going down but it's still too high for a "fast" model

So about the same “intelligence” as Muse Spark 1.1 but 2x faster and about 2x as expensive.

Google is walking backwards, with such a pile of cash in pocket, i feel they are doomed.

Nice to see that it's cheaper than 3.5

Models are expensive and low performance. On top of that they make you jump through hoops to even use these models without being throttled even for the weaker models. The only reason we are using them is credits. As soon as credits run out we are switching immediately.

Haven't been excited for a Gemini release since December. Wild to see.

I'm going to get downvoted/flagged but I feel like we need a new type of "Show HN/Tell HN" etc for "New AI Model Available".

Front page is tedious these days.

the real product is the naming confusion we made along the way. Gemini 3.6 Flash, 3.5 Flash-Lite, 3.5 Flash Cyber — at this point even the model cards need a model to explain them

Not good enough for high-end, not cheap enough to be for low-end. Next!

quite a good model, the speed/price/quality ration is a new golden intersection for me, not sure if its as good as grok 4.5 but quite fast/capable model.

It feels like AI is going to be the end of Google. The post-Schmidt company culture cannot produce consistent, consumer-friendly products that any sane person would want to use consistently.

I have no skin in this game and this comment will be gray in a few minutes BUT a friendly reminder that these types of threads are astroturfed heavily by competitor labs and any info should be taken with a massive grain of salt.

"The model will be exclusively available to governments and trusted partners via CodeMender soon as part of a limited-access pilot program"

we are stealing plutocracy from the jaws of emancipation.

i don't want to live in a world where abundance is guarded and shared among politicians and cronies, whilst the rest are left to rot.

never tried gemini for coding, but this news seems to be compelling, i would definitely give it a try

dupe? https://news.ycombinator.com/item?id=48993130

2 red flags

1- no comparison with gemini 3.1 pro

2- no comparison with any other model

Specific to task these can be huge plus point.

no actual cyber model release, useless

If only I could use Pi.

I remember back when Gemini looked like it was the best model that this comment section was full of confident predictions that Google had "won" and that no one would ever catch up with them again. The most embarassing part is that I kinda believed them.

I think it’s safe to say Google seems a bit out of the top AI competition now. The “cyber” stuff also starts to become laughable with open models providing the full power without the crap Anthropic, Google, OpenAI are trying to frontload on you(I.e you are not allowed to develop/review a login system, pay a special cyber operation team to do it for you). They really deserve to become irrelevant in the future of AI.

if only DeepSeeek supported vision, would never use Gemini.

Other discussion from a few minutes earlier: https://news.ycombinator.com/item?id=48993130

3.6 Flash has the same performance on artificial analysis benchmarks as 3.5 Flash. So... what... is... the... point?..

The silence is deafening.

Google watches over the last few months a flat out assault on the Pareto curve from American and Chinese companies. Release after release pushing the boundaries of frontier intelligence and price/performance.

And the response from arguably the biggest AI research labs in the world by headcount is Flash 3.6.

What do you do when you are given essentially unlimited resources and still find yourself falling behind?

Plugged 3.5 Flash Lite into an existing agent harness that was previously using 3.1 Flash Lite and this shit just does not work. It's not following instructions and is not producing the correct tool calls.

Gemini 3.5 flash is already a pretty good model. But, unfortunately, the primary way you can interact with it for coding is through Antigravity - which is actively developer hostile.

It doesn't matter how good the model is if you're (mostly) forced to use it in Antigravity - which turns any model into crap.

Wake me up when Antigravity doesn't suck.

Why exactly are they announcing these completely milquetoast models ?

I'd be low-keying the release if anything, given how lame they are compared to their competition.

What am I missing?

I use 3.5 flash 10x more than any other model, despite have access to all of them. If I'm going to play a slot machine, I'd rather get the pain over with quickly.

We're almost five years into the whole GenAI thing and we're still relying on these guys to spoonfeed us incremental updates.

It's time for them to start focusing on open-weight models and efficiency. Otherwise there's just a layer of marketing hype and "will it do this?" that has to be cut through for evaluation of each and every release cycle.

Models are getting easier and easier to create. The money, if there's any here, is in the harness the user interfaces with, and the data centers running them.

Google seems to be falling way behind the pack. antigravity cli is pure trash. gpt 3.5 pro is now behind and isn't released yet. GPT 6 and Fable 6 releasing next month. What the hell is going on over there ?

Just switched to AI Plus from Pro, seems like I won't be missing much.

> we have taken an intentional approach to deploying 3.5 Flash Cyber. The model will be exclusively available to governments and trusted partners

Screw your government! US and Israeli governments should get the least access, but of course we all know they'll be the (only) ones to get full unfiltered access.

3.5 Pro must really suck.

They are comparing against their own previous models instead of competitors. Not a great sign.

At this point just put the Pareto in the bag bruh

whatever, dude. give gemma5

is it just me or is this one-upping each other every few days getting ridiculous secreting a whiff of desperation?

Some more discussion:

Gemini 3.6 Flash https://news.ycombinator.com/item?id=48993130

I keep saying this and people dont believe me, but I have b2b saas systems with actual agents running around the clock, and the performance/stability of the flash model is higher than most other models.

Meaning, its predictable with tool calls, wont spin off a million tools/do weird behavior, its reasonable. Even sonnet in a real world decision making scenario is not reliable, or will reason so long its incredibly expensive.

The benchmarks arent catching all the value, and most people have never actually ran an ai agent in a real context that matters

With all the naysayers on Gemini models I'm curious how many people actually use Gemini regularly?

For me, Gemini models are the most usable. Claude Opus and Mistral always try to turn queries into one-shot enormous commits, which just burns tokens, time and annoys me for something which is still wrong more often than not.

Gemini seems far better at listening to instructions and giving me what I actually want, on top of using far fewer tokens and wasting my time. Fable is the only model that's come close to Gemini Pro for me.

And as this is about Flash, it's exciting, I find Flash can usually get the right answer pretty quickly and without too much nonsense.

not a google fanboy by any stretch... though i've been thrilled with the flash line of models... i exclusively use it on high, and have found it to be a great fit for increasing productivity 10-fold while maintaining quality... sure it can't just go off and one-shot a bunch of work, but at the complexity level i tend to work at, neither can the frontier in a robust way that i can be confident in... sure i have to be in the loop more, but that helps keep me grounded and course-correct earlier before wasting tokens... and when you sufficiently spec out a coding/software problem, and i mean really document all of the critical nuance, it will successfully satisfy the constraints... the quality is rarely acceptable on first-pass, but it forces me to stay connected to the architecture more than i would be if using a frontier model... i've found this to be a happy middle-ground of productivity and awareness...

[dead]

[dead]

[flagged]

[flagged]

[dead]

[flagged]

[dead]

[dead]

tested the models on aistudio. despite that the knowledge cut off is march 2026 it still knows nothing about 2025!

you can check by asking "list notable world events in 2025, only list unplanned" on aistudio. or you can ask for Charlie Kirk, it also does not know. I tried it multiple time to ensure that I didn't not get routed to older models!

> but google has search

irrelevant, without deeper knowledge about cutting edge technologies or latest libraries, all of it suggestions are crap. even you ask it to search it will still use outdated keyword thus only getting outdated information.

in other word, what a disaster!

Google desperately needs to make some leadership changes within their Gemini team now that they've been surpassed by 3-5 open weight models and risk loosing frontier status all together in the near future.

At this point, I think google should consider becoming a hyper scaler for anthropic and open ai, and I predict that that is exactly what they do. The model is no longer the most valuable part of the stack.