experimenting w/ flash 3.7 on morphology and it looks promising. google models looks capable on linguistic side and prose. are there any reliable benchmarks on this?
Gemini 3.7 Flash
471–480 of 525 posts
Re: Gemini 3.7 Flash
#472Earlier quoted context omitted.
Sergey's out spending $100 million to lobby against the billionaire tax in California. Lest his net worth shrink from roughly $250 billion to $237 billion. Can't become a pleb now! [0] [0] https://techcrunch.com/2026/08/10/google-co-founder-sergey-b...
not that I'm intending to defend Sergey here but, if you could spend 0.04% of your net worth to protect 5% of your net worth (and probably all of your easy liquidity)... wouldn't you?
No. It's irrational at that level of wealth. He can make it back in a year. Billionaires very commonly see 20% returns a year.
I'm not in favor of the billionaire tax, there's better ways, but if the people want it he can just put his big boy pants on and fucking deal.
Re: Gemini 3.7 Flash
#473It's just useless to build any model in this range because deepseek flash is so fast and so cheap. I see no reason for anyone to not use it for a few points in performance. Only models that matter are the edge fable class models people use for code, and Google struggled with that.
Rates are going up Aug 16.
Re: Gemini 3.7 Flash
#474Earlier quoted context omitted.
Also best at OpenSCAD, seemingly for the same reason, at least in terms of "iterate on a design, comparing visual output to target".
Are you manually rendering previews of its OpenSCAD output to create images for it to review, or do you have a workflow that automates that?
I've built some tools using AI. A custom GUI that has "copy context" and "copy image" buttons, to make prompting easy, and the camera position is persisted to disk on each change so that the agent can run a command to get a screenshot of it.
I'm thinking about inlining an AI chat window directly - I guess I might fire off a prompt to do that right now.
Re: Gemini 3.7 Flash
#475Earlier quoted context omitted.
Also best at OpenSCAD, seemingly for the same reason, at least in terms of "iterate on a design, comparing visual output to target".
I've found Sol excellent at OpenSCAD. But I don't do "compare visual output to target" as much as 3D reasoning type tasks (eg: "Build a G1 curve where the -X face meets the +Z face" etc)
"G1 curve"? Nice, cool to have new vocabulary for the spell book. Thanks.
Re: Gemini 3.7 Flash
#476This hits different after a long week
Re: Gemini 3.7 Flash
#477Earlier quoted context omitted.
+1 GCP can be confusing - I totally get that. Even if you have GCP projects, I'd still recommend the AI studio UI - easier to figure out. Also, you can easily see the free tier in AI studio and just use your API key from there.
Yes I will look into it now but there’s no cross reference from GCP, and why are there two ways? And why did the GCP way require all these APIs enabled while AI studio didn’t?
> And why did the GCP way require all these APIs enabled while AI studio didn’t?
Because whatever internal team owns AI Studio fought for approvals to do so and GCP didn't?
Re: Gemini 3.7 Flash
#478Earlier quoted context omitted.
Fable orchestrating DeepSeek v4 Flash to implement a plan is my new favorite thing. It's so freaking fast, but you gotta tell Fable to watch Deepseek like a hawk or it'll go off the rails.
Interesting - how are you interacting and orchestrating this?
Not the parent, but:
You define roles for different agents like this:
modelRoles:
task: fireworks/kimi-k3-fast:high
plan: fireworks/kimi-k3-fast:max
slow: fireworks/kimi-k3-fast:max
smol: fireworks/deepseek-v4-flash-0731:low
tiny: fireworks/gpt-oss-20b
vision: fireworks/qwen3.7-plus:high
designer: fireworks/qwen3.7-plus:high
advisor: openai-codex/gpt-5.6-sol:high
main_worker: fireworks/kimi-k3-fast:high
fast_worker: fireworks/deepseek-v4-flash-0731:low
vision_worker: fireworks/qwen3.7-plus:high
research_worker: fireworks/glm-5.2:medium
code_worker: fireworks/kimi-k2.7-code-fast:high
review_worker: anthropic/claude-fable-5:high
security_review_worker: fireworks/kimi-k3-fast:max
minimal_worker: fireworks/gpt-oss-20b
default: fireworks/kimi-k3-fast
task:
agentModelOverrides:
task: "@main_worker"
sonic: "@fast_worker"
scout: "@fast_worker"
designer: "@vision_worker"
librarian: "@research_worker"
reviewer: "@review_worker"
security-reviewer: "@security_review_worker"
Re: Gemini 3.7 Flash
#479Here's a image->html test. Gemini has always swung above its weight class for vision work, so I'm always eager to try it with this. Original images: https://image.non.io/neonRamenDesigns.webp Gemini 3.7 build: https://html.non.io/neonRamenGemini3.7 Opus 5 build for comparison: https://html.non.io/neonRamen Opus is still best in class for this, but it's worth noting how well Gemini 3.7 does vs a more comparable LLM pr…
These look exactly like all of the low budget bodega signs near me. They also look like a bunch of cheap ads for parties that I keep seeing. The sameness of style is uncanny. (I don't have the Bodega signs, but I'm thinking of shit like this, from a quick google: https://linkstub.com/en/wet-wild-foam-party )
All AI "art" is proving out to be basically crap, and the good thing is that it has started to become a very good filter, as in whoever uses AI art most definitely is doing a shitty job with the rest of his/her business so it's better not to bother with said business.
The thing is though that most of the nerds here are oblivious to all that, so it will take a while for them to fully acknowledge what's happening. In a positive note, I'm here on a AI thread discussing AI-image generation and I haven't yet seen any link to the dreaded pelican on a bicycle thing, so hopefully we're getting into the right direction.
Re: Gemini 3.7 Flash
#480I just subbed to Gemini a week ago and have been using antigravity and 3.6 flash. The speed is absolutely a differentiator compared to Claude.
Claude is terrible when it comes to speed.
Flash is great, but Codex models are also fast, as is DeepSeek v4 Flash.
Anyone who's Anthropic-pilled should really get out and explore and see how unbelievably terrible they are when it comes to speed and cost vs quality.
Anthropic has good models, they're just way too expensive and slow for what you pay for.