Hey everyone 👋
If you’ve been building in Nowa over the last ten days, you’ve probably noticed that the agent handles longer and more complex tasks, fixes bugs that used to trip it up, and produces more unique, polished designs with animations. And you’ve definitely noticed that you can use the AI a lot more every day, on every plan.
Here’s the full story of what we changed to make that happen.
We switched to Gemini 3.6 Flash the day it launched
On July 21, Google released Gemini 3.6 Flash, the successor to 3.5 Flash — and we moved Nowa over to it the moment it went live. This is one of those rare upgrades where the new model is better on every axis, and here’s what that means for you in practice:
It’s noticeably better at coding and debugging. On DeepSWE, a benchmark that measures whether a model can fix real software bugs on its own, 3.6 Flash jumped from 37% to 49%. Just as importantly, Google reports fewer unwanted code edits and fewer failed retry loops — so when you ask the AI to fix something, it fixes that thing instead of rewriting half your screen.
Long tasks stopped being a struggle. The new model completes multi-step work in fewer steps and fewer tool calls. In independent testing, the average time to complete a task dropped from 2.7 minutes to 1.3 minutes. Combined with its 1-million-token context window, this is why building a big, multi-screen feature in a single prompt now works so much more reliably in Nowa.
It’s better at understanding and generating UI. Google specifically improved multi-element layout generation and converting visual references into working screens — which is exactly the kind of work Nowa’s AI does all day.
And it’s cheaper to run. Google lowered the model’s pricing compared to 3.5 Flash, and the model itself is more efficient: it uses about 17% fewer tokens to complete the same work — and on some coding tasks, up to 65% fewer. Put the lower price and the efficiency together, and the same task now costs roughly a third less than before, with even bigger savings on heavy coding work. We’re passing that straight to you: every credit you spend in Nowa now simply builds more.
Instant mode got a serious brain upgrade too
We also upgraded the model behind instant mode, from Flash-Lite 3.1 to the new Flash-Lite 3.5.
We know a lot of you live in instant mode for quick design changes — tweak a color, adjust a layout, polish a component — because it responds in seconds. The new Flash-Lite keeps that same near-instant speed, but it’s dramatically better at code: on terminal-based coding tasks it scores 54% versus 31% for the old model, and it even outperforms some bigger previous-generation models on software engineering benchmarks. It costs us a little more per token than the old one, but the quality jump is worth every cent: your quick edits now come out looking like they took much longer than a few seconds.
We rebuilt parts of the agent itself
A better model is only half the story — the agent around it matters just as much. Over the past weeks we’ve shipped a long list of improvements to how Nowa’s agent plans and executes complex tasks: breaking big requests into the right steps, recovering from errors instead of spinning on them, and staying consistent across long builds.
We also tuned it heavily for modern UI. The agent is now much better at producing polished, contemporary designs with smooth animations — the kind of details that make an app feel finished, not generated.
Double (and triple) the daily AI credits
Here’s the one you’ve all noticed. We’ve been watching what you build in Nowa, and honestly, the apps keep getting more ambitious — real products with real complexity. Ambitious apps consume more credits, and we didn’t want limits to be the reason you stop mid-build.
So we’ve doubled the daily AI credits on Starter and Launch, and tripled them on Scale:
| Plan | Daily AI credits |
|---|---|
| Starter (free) | $5 / day |
| Launch | $20 / day — 4× Starter |
| Scale | $50 / day — 2.5× Launch |
Your credits still reset every single day, and plan prices haven’t changed. Combined with the cheaper, more efficient model above, the amount you can actually build per day has gone up a lot more than 2–3×.
Try it all for free this weekend — August 8 & 9
To make sure everyone gets to feel the difference, we’re bringing back the free weekend: on August 8 and 9, AI and deployment are open for everyone, no limits.
This is your moment to build the serious app you’ve been putting off — with all the detail you want — and to take Nowa 3.10 and everything above for a proper spin.
Start building →
We’re here if you need us
Questions, feedback, or something not working the way you expect? Email us at team@nowa.dev, or come hang out with other builders on Discord and Reddit — we’re always around.
Happy building! 🚀
— The Nowa team
