Flux + GPT-5 (Beta)

In the right scenarios, it’s already delivering sharper reasoning, smarter reviews, and more accurate design decisions than anything we’ve shipped before. We wanted to get it into your hands immediately so you can explore what’s possible alongside us. It’s early, it’s raw, and we want you to push it. Break it. Tell us where it shines.

Try It Now

You can start using it right away. Open any project in Flux and launch Copilot. Click the model dropdown at the top of the chat panel, select “Next-gen” and then give it a real challenge. Some great starter prompts to see its strengths include:

“Perform a top-to-bottom schematic review for correctness, completeness, and robustness. Assess power, clocks/resets, signal interfaces, analog paths, protection, and passive choices.”

“Replace all low-stock parts with alternatives that meet the same constraints.”

Try GPT-5 in Flux now

What’s New and Better

The upgrade isn’t just that GPT-5 is a newer model. It brings a different caliber of intelligence to Copilot:

• Stronger reasoning and planning for complex, multi-step problems.
• More accurate part and constraint decisions, often with clearer explanations.
• Sharper design reviews that catch subtle issues and propose fixes.
• Richer, more verbose answers that lay out assumptions, tradeoffs, and edge cases.

These improvements land harder in Flux because Copilot already has deep, live context on your design—down to parts, pins, nets, properties, constraints, and stackups—so reinforcement models and LLMs can work side-by-side from the canvas up to system architecture. And because Flux is built for agentic workflows—stepwise actions, constraint-aware edits, and iterative design loops right where you work—GPT-5 isn’t starting from scratch; it applies improved reasoning directly to your schematic or layout. Layered on top is a knowledge base of industry best practices and embedded design/process checks, so your AI partner starts from seasoned experience and turns that context into answers that are immediately relevant and actionable.

Early Win from the Weekend

In just 48 hours of testing, we saw moments that made us stop and say, “This is new.”

Design a low-noise microphone preamplifier for an electret condenser mic feeding a 24-bit ADC. You must calculate the bias network, gain-setting resistors, coupling capacitors, input high-pass cutoff, output anti-aliasing RC, and decoupling layout. Follow the op-amp and microphone capsule datasheets, ADC input requirements, and industry best practices. It will be integrated into a design. Supply: 3.3V analog rail. Mic bias: 2.0 V through resistor, current ~0.5 mA. Target gain: 20 dB to 40 dB switchable. Bandwidth: 20 Hz to 20 kHz. Input noise target: as low as practical. Include pop-suppression considerations and star-grounding strategy.

Try it Now

In this case Flux took a plain-English prompt and produced a full low-noise mic preamp to a 24-bit ADC—calculating the right bias, gain, and filter values, choosing real parts, then placing and wiring the entire block with decoupling, VCM bias, and star-ground best practices. It even audited itself (fixed missed ties, made gain legs switchable). The result is a ready-to-review schematic 80% away from layout built end-to-end—complex, competent, and fast.

Getting the most from “Nex Gen” model in Flux

• Read the whole answer. The “Next Gen” model is verbose on purpose—the assumptions, edge cases, and self-checks are where the value lives.
• Front-load context. In your first message, share goals; rails & loads; key interfaces (e.g., USB-C: CC1/CC2, D+/D−, SBU); constraints (cost/size/EMI); and manufacturing rules of thumb. Keep answering follow-ups to deepen the design.
• Scope tightly. When wiring schematics, tackle one rail, one bus, or one block at a time.
• Plan before you act. Validate the change plan against requirements before applying edits—changing the plan is cheaper than undoing work.

What’s Next

Right now GPT-5 powers Copilot’s chat, but this is just the beginning. We’re already working on:

• Tighter loops between chat and in-editor actions.
• More constraint-aware placement and routing.
• Datasheet-to-design transformations in minutes.
• Smarter, in-context automated fixes.

Open Flux now, switch Copilot to “Next-gen” and see how it handles your next design challenge. The sooner you try it, the more your feedback can shape the next leap in AI-powered hardware design.

Try GPT-5 in Flux now

‍