I made https://jsmess.com as one of my first vibe code projects and have been incrementally improving it over time when I have leftover tokens...
It's closest to JSFiddle but completely static, hosted on GitHub pages. No server side processing. Projects are private, saved on localstorage and can be shared via encoded URLs.
This has been a really fun way to ramp up with LLM based coding and I learned a ton, and made a tool I actually use quite often.
Can you elaborate on running them "in conjunction"... are you running the same query on multiple models and then using a third model to judge or make consensus? or am I misunderstanding completely. I'd like to understand how these small models "run together"
Late, but it depends. Sometimes I ask for all of them to do the same thing and have a different model judge it, sometimes I ask for an orchestration model to run the smaller ones each on one task. I have a 'model cohort' where I run the same request across 5 very inexpensive models. Then have my driving LLM judge or synthesize.
You just come up with what you want in your head, and tell it to do it in that way and it does it. I trust 5 independent smartest programs ever over the SOTA smartest program's only.
I do this in OMP, a fork of Pi. It lets you set different models for different tasks. So with an API that has many different companies models I can set the Plan model to the best one, right now I am using GLM 5.2 for that, it plans really well. I have Vision set to Kimi 2.7 Code (cheaper and vision is just fine). Minimax M3 is set to the Advisor role (double checks work). Deepseek v4 Flash is set for the Task role. And MiMo 2.5 pro is set as default.
With this setup GLM handles planning and managing my AGENTS.md, and orchestrating subagents for tasks from the plan/todo GLM created. The tasks themselves are handed off to Deepseek v4 flash to implement with strong instructions and examples for each agent. Minimax M3 reviews the output as the Advisor and recommends changes, catches bugs, and whatnot, subagents can be re-run with that information.
Overall I am saving a lot using some of these smaller models. But with this setup I am getting great results.
Stock OMP. When you add a provider and then /model you can choose a model and choose what roles it will take. Do that for each role it has.
Then I use /plan when I need to be planning and not writing to any files. Take the advice the UI gives you where it tells you to add the word orchestrate into your prompts when you want to make sure it uses its todo/tasks and sub-agents.
I opened the tab in the background and when i got around to it later somehow i lost the game.... i think you need an official "start" button to prevent this.
Wow. As a comparison, I just opened a new Google Maps tab in Chrome.
According to the Chrome Task Manager, the tab alone uses 433mb RAM and 34mb GPU memory footprint after first load.
Florb (an FLTK based map viewer similar to Dillo in specs) would use almost the same RAM as native Mac OS 9 software. And yet people say "JS software it's far better". Even a TCL/TK based viewer with TCCLib/TKLib it's far lighter.
JS is easier to develop, easier to deploy cross-platform, and, more importantly from a business perspective, easier to hire for. So it makes complete sense to use JS in the commercial world.
I'd like to study your setup. Would you be willing to share?
Perhaps a github repo of your 5 extensions or even a pastebin if you would be so inclined. I would be grateful to learn more about this by studying from your success...
I might share it at some point but I think it's quite similar to a lot of others out there, except that it's very specific to my personal projects and goals. If I shared it I'd need to spend at least a while cleaning up and improving docs.
It's one of the reasons I suggest you study the famous setups (oh my pi, or superhuman skills etc.) and convert them to your personal needs.
Another hot take from him in 2018 is "Many people I respect here, lately identified Facebook as the root of all the evil. I want to start this thread about why I disagree..."
Per the "Availability" section of the page, seems like should come back to all plans eventually...
* From today through June 22, Fable 5 is included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost.
* On June 23, we’ll remove Fable 5 from those plans. Using it after that will require usage credits. If capacity allows, we’ll extend the included window.
* After this point—when sufficient capacity allows us to do so—we aim to restore Fable 5 as a standard part of subscription plans. We intend to do this as quickly as we can.
Coding plans are a (massive) subsidy. We can debate until the cows come home whether western frontier models' API pricing rates are fair, but the coding plans are all heavy discounts below those API rates meant to draw people in and get them hooked (and, ostensibly, to be useful for hobbyists or other lower-usage cases).
It's been discussed at length (on this site, on other sites, on like every blog ever, etc) that, eventually, those subsidies will end, much as the $5-10 Ubers/Lyfts I used to take from the far north end of Chicago into the Loop in 2016 would eventually end once those companies had a footing and didn't need to hook folks.
So - yeah, I mean, a v5 model launching in a year where Anthropic has a rather deeply established market and in a year where AI costs are rising from nearly all providers (sometimes for multiple reasons) seems like exactly the thing I'd expect them to pull the subsidy plug on after a launch teaser.
(Even the open-weight models sometimes do this: for example, OpenCode Zen/Go has a rotating door of free models at any given time that eventually leave the free tier and move into the paid tier once the launch day hype/marketing dies down)
The worst part is that Uber "only" lost about $30bn. AI will probably lose at least $300bn by the time the bubble pops. Which means that the pressure to hook and enshittify will be at least 10x as high.
Problem with that website/perspective is separating training costs from inference costs. Training is a one time cost, and while it is certainly not something you can completely ignore, it being one time changes the answer to "Is AI profitable?".
That site doesn't list the dozens of companies doing pure inference, and making a profit while doing so.
What's Google's business case for releasing open models? Don't get me wrong, I am grateful and appreciative of these releases. I'm trying to understand how it fits into their bigger picture as a for profit company? Are they not helping competitors build on the novel technology they have developed?
Is it simply goodwill and/or marketing? Or am I missing something strategic?
A big part of the frontier labs abilities to charge 80% gross margins on inference is having the cornered resource of frontier models.
If that inference becomes popular and valuable enough that those companies make billions of dollars in profit, those companies could use that profit to fund the building of alternative products and platforms that dis-intermediate google's relationship with the customer.
Google already has an 80% gross margin business, the biggest one in the world. Everybody wants a slice of it.
By offering frontier inference closer to cost and open-sourcing everything that's sub-frontier, they're commoditizing frontier labs' models, which inhibits their ability to durably make high gross margins on inference.
A 12B-sized model is a far cry from "frontier inference". That's more like DeepSeek V4 Pro territory which is a 1.6T model. Or for multi-modal models, Kimi 2.6 which is 1T.
2. I'm not sure what this has to do with the case, unless you're arguing Google has an ads monopoly, in which case the best argument would likely not be that adwords lead to bidding wars because that just sounds like they're selling a product people really want to pay for
3. There's nothing criminal about being a very diversified business
You're right that it's not literally frontier. But like recent Qwen releases, it is a lot more capable than anybody thought models of this size could be a year ago, like capable enough to set a ceiling on what you can charge for AI for certain applications. Others still clearly justify a stronger model, but this trend may continue, etc.
Basically with upcoming spark laptops, the smaller models will likely get fine tuned to interface with google services. Then, Google can essentially make Chromebook software include those models, which is the same use case as android.
And you better believe that they will be collecting user data and building advertising models.
This won't replace commercially viable, revenue generating alternatives of their own devising, but it does enable development activity and initiate conversations with enterprises who start with this model but want to do slightly more.
That's my experience right now... my company is all in on a plethora of platform products. Also, Microsoft just yesterday said their goal was "Unmetered intelligence". There's a lot of things that can be enabled by small local models, and those things are part of stacks that can generate revenue in other layers.
Could say the same for camera processing in the Pixel Camera app or any other binary someone wants to re-use that comes included in a software distribution (seemingly for 'free'). They can't lock the instructions up on the server so they might as well make the binary be freely distributable?
Companies don't commonly give away executable binaries "just because", why'd they start now for these binary blobs that are the models?
Not that I'm unhappy about it! Yay for open data any day, I'm just not understanding why, at least beyond PR in nerd circles
Binaries are source code outputs, they are copyrightable and patentable. Weights are not copyrightable so people can freely extract the weights and run them. If Google patents any of the novel algorithms here releasing it all freely isn't an impediment to making people license it.
Are you sure that isn't about LLMs' outputs? There I know there have been some court cases that say this, but the model itself is a work created in intricate and somewhat creative ways (I hesitate to use the word "creative" here, but would similarly hesitate to label a routine picture of the moon creative whereas pictures basically always have copyright; the bar for creativity is basically an epsilon amount above zero, afaik)
Because a model like this can't be as easily obfuscated as image processing. Image processing is a bundle of many moving parts, a lot of functions each with it's own inputs and outputs. A model is a single function which can be easily extracted and reused, in comparison
Arguably, but that's not the point. Take image (e.g. png) files on a CD-ROM shipped by a game vendor, which can be trivially copied even by my grandma. That doesn't move the game vendor to release them as freely distributable under the Apache license
Good point but still, why would Google police this model? If they had a restrictive licence on it do you think it would be worth it for them to enforce it? This way they at least buy some good will and mindshare
That makes sense to me. Guess one might say the same for game icons and other such files that lay around in disks, but yeah maybe it's as simple as that
Not quite the same, understandably Blizzard cares a lot about their IP because otherwise private servers leech their users. Maybe a small game designer cares a lot about the small game they made or whatever since that's all they have. A four trillion market cap company can afford to be "charitable".. where it costs them nothing and might cost them more to enforce their rights.
They could lock them down legally which would prevent commercial use, but they choose not to, and they boast about how many tens of millions of times Gemma models have been downloaded by developers.
So there must be more to the rationale than just local model weights getting hacked out of devices.
Neutering OpenAI and Anthropic would be my guess. Commoditized LLMs won't hurt Google nearly as much as it hurts the LLM-only companies, and so accelerating the inevitable just helps knock out potential future competition in areas where Google -does- make a lot of money now.
I sincerely wonder why. Chinese censorship is only really relevant if you're doing anti China stuff, which is to say never, while the Western kind of model censorship ( a combination of copyrights and general fairness ) are something everyone's had to work around at least once, even if just for writing an interesting story.
It’s about enterprises who care about supply chain risk and having a throat to choke if they have a problem.
Here’s a real example.
I’m in a design meeting talking about a model use case. We have a question about the data pipeline or the prompt format that would benefit from knowing about how the model was trained. The enterprise team lead calls the dev tech engineer from the company who produced the model. He is already in the office and walks into the meeting to answer the question.
Google is one of the few verticalized options in AI: Data, models, cloud services, low-level silicon (TPUs), internal use cases, retail use cases, B2B uses, distribution (browser & mobile), etc.
They rise with the tide of AI adoption. But they gain ground if people opt into Google solutions. And any token sent to a Google model (free or paid) actively punishes their competitors that are then required to spend vast sums to remain bleeding edge.
If you're an AI lab, you definitely want research teams in this space - as this is where you can most easily iterate and make improvements which you'll then bake into larger, frontier models.
The question is: do you want to release your models, or use them purely for R&D?
Since everyone else is already releasing models of similar qualities, it's hard to say you're shooting yourself in the foot if you join the chorus.
The added cannibalization of releasing them is effectively zero, so the reputational benefits are likely to be worth it.
>The added cannibalization of releasing them is effectively zero, so the reputational benefits are likely to be worth it.
Nobody would be looking at Qwen if their ~30b class models weren't fantastically good, it's great advertising and builds significant goodwill with developers, who are going to be your biggest advocates.
The other thing is, all these models are already disposable grade, and in a year they'll all be outclassed by The Next Big Thing. "Open" models are less than 18 months behind SOTA right now and I can't imagine that will slow down much over the next two years, they may even begin to close the gap. Nobody even talks about llama 4 anymore despite only being a year old.
As long as Chinese firms are releasing good open models I imagine there isn't a huge downside for Google to release state of the art small models to compete in the "free" space.
It's to destroy possible footholds for competitors and prevent them from making money in segments that Google doesn't care too much about, but can trivially commoditize.
I think its even more puzzling because you can't even run Gemma 31b on google cloud, they only let you test it with a rate limit. No way (I can find) to actually pay them to use it.
We saw great results in our usecase using google direct. Moved to Openrouter because google wouldn't let us use it beyond a test.
Then Openrouters performance looked worse, not sure if there was a quantized version or something. So we instead looked at Deepseek v4 Flash, and opted to go for that.
This model would probably be great for a super low cost cloud model, would love to use it in the cloud, Google makes you go elsewhere.
It’s on openrouter. We just noticed performance was worse in a specific agentic app usecase. It’s possible we made an implementation mistake, my main point though is Google is really silly not hosting their own models.
I tested Gemma 4 31b for OCR and it's very good at it. This makes sense because I also get the best OCR results from Gemini compared to Claude or ChatGPT in my use case.
Google's MO since always has been to release great products or services for free, position themselves high and then abandon them or just find uses for Enterprise sales.
I'm pretty sure they are doing it because they get some research experience by shrinking and improving these models, and because they know that by doing this they get some good PR among the dev community.
Google's "free" is and was ad-supported, even if some products now have a paid tier. These models don't include ads. Doesn't seem like the same underlying reason
A strong business case for Gemma includes fine tuning, adding AI to apps that run in the cloud, strengthening Android, shifting unprofitable small AI compute to devices, and harming competitors. The first two would be done using Google's cloud services due to integration with Gemma. I think Google is currently the best positioned company to profit from AI sales to businesses over the next few years, and Gemma is a critical part of the story.
Google is actively, and directly helping companies continuously train use-case specific models based on Gemma 4 foundation. The company gets a model they fully own, trained on internal, sensitive data, and Google scoops up the profits from the training and ongoing compute spend to keep the model up-to-date.
Isn't Apple about to license some variation of this from google for on-device AI? Maybe it’s their sales pitch to Apple and then they will lock it down.
The complete Chinese worldwide domination in this sector would be the alternative, since nobody else is releasing anything meaningful.
Plus every open model undermines their local competition by furthering open research and reduces moats, especially since Gemini as a frontier model isn't really competitive with GPT nor Claude for most applications.
Maybe they are hedging against a future where local models are just as good as cloud models? Or maybe they can go the Taalas route and start hardcoding Gemma on a chip and hardware manufacturers can use it for local private AI.
Competition from Chinese alternatives hopefully forces more openness and efficient models. DeepSeek for example is nearly on par and far more resource efficient, good for the planet imo
It's closest to JSFiddle but completely static, hosted on GitHub pages. No server side processing. Projects are private, saved on localstorage and can be shared via encoded URLs.
This has been a really fun way to ramp up with LLM based coding and I learned a ton, and made a tool I actually use quite often.