Gemini 3.5 Pro Is Coming to Felo AI — Free Access to Google's Newest Model
Google's Gemini 3.5 Pro delivers 92% of GPT-5.5's performance at 1/15th the cost. It's arriving soon on Felo AI with free access. Here's everything you need to know.
Google skipped 3.4. Went straight to 3.5.
Companies don't do that for a routine update. They do it when the new thing is different enough that the old numbering scheme doesn't make sense anymore.
We're adding it to Felo AI soon. Free from day one, like everything else on the platform.
Here's what makes it interesting.
What Is Gemini 3.5 Pro?
It's Google's newest model, announced at Google I/O 2026. And the numbers coming out of it are worth a closer look.
On coding and reasoning benchmarks, Gemini 3.5 Pro scores at 92% of GPT-5.5. The context around that matters just as much as the number itself. GPT-5.5 costs fifteen times more to run.
So you're getting almost the same output quality at a fraction of the price. That's the story here.
Why It Matters
Speed
You type something and wait. That pause is where most AI interactions lose their momentum. Gemini 3.5 Pro answers most queries in under 200 milliseconds using distillation and sparsity to keep the capability while cutting the latency. The whole thing feels closer to a conversation than a command queue.
Coding
People who've tested it are generating complete applications from single prompts. Not snippets. Full things. A Windows-style web OS. A Minecraft clone. An MMORPG with over 4,100 lines of code. On benchmarks it hits 92% of GPT-5.5's coding score. At 1/15th the price, that ratio starts changing how teams think about which models to run.
Cost
The leaked pricing is $0.25 per million input tokens and $2 per million output tokens. GPT-5.5 runs fifteen to twenty times higher for similar quality. The cost difference is large enough that it matters for anyone running models at volume.
Multimodal
Text, images, multi-step reasoning, all in one model. You can upload a chart and ask what it shows, describe an image, or walk through a problem with several moving parts. No switching between models to handle different input types.
Who This Is For
Developers can use it for single-prompt app generation with a fast enough feedback loop to feel like actual pair programming. Researchers can analyze papers and synthesize findings with a knowledge cutoff that reaches early 2026. Content creators spend less time waiting on responses and more time editing. Data analysts can work through tables and charts without losing context partway through. Students get free access on Felo with no cost barrier. Business teams can run high-volume tasks like report drafting and document analysis at a price that actually makes sense.
How It'll Work on Felo AI
Two options when it goes live.
Felo AI Search lets you select Gemini 3.5 Pro as your model and ask questions directly. Search results come back with AI-powered answers and citations.
Felo LLM Playground lets you pick the model manually and chat. You can also line it up against other models and compare outputs side by side.
Both free. No credit card.
Coming Soon
Google didn't iterate on this one. They jumped a version number. And they priced it at a level that makes everyday use realistic instead of reserved for special occasions.
We're putting it on Felo AI as soon as it's ready.
Try Gemini 3.5 Pro on Felo AI when it goes live
This post is also available in 简体中文, 日本語, 한국어, 繁體中文, हिन्दी, Français, العربية, Русский, اردو, Bahasa Indonesia, Deutsch, Tiếng Việt, Türkçe, Italiano, ไทย, Español, বাংলা and Português.