Faster results and steadier sessions

Ampere now returns AI Tool results sooner, keeps more parallel work moving, and holds onto session, browser, and saved-work context more reliably. This release also refreshes account setup and updates the lower-cost model option to DeepSeek V4.1 Flash.

What's new

• AI Tool results return sooner. Completed runs now come back through a direct, durable handoff instead of routine status polling and fixed waits. Ampere can continue the original response as soon as the committed result is ready, while exceptional recovery remains available if the connection is interrupted.
• A clearer start for new and returning users. Sign-up, sign-in, verification, and password recovery now share a cleaner responsive experience with more useful inline guidance. Guided Brand Context setup is now a two-pane studio: you can follow research and build activity, watch your Brand Canvas fill in, preview saved context, and confirm competitor domains before continuing.
• Sessions become recognizable faster. A useful title appears as soon as the first message is accepted, then may be refined while Ampere is still working. An explicit title you choose remains in control.
• Saved work and confirmed actions stay connected to the conversation. Saving or updating a reusable Project Asset now adds a card beside the response, and asking Ampere to show an existing asset brings its current card into chat. Confirmed Monitor actions and other required-input steps are better able to show their saved outcome and continue after a refresh or runtime restart, without repeating completed work.
• Browser work is more dependable. Local and remote browser sessions can keep the intended browser, tab, downloads, and recordings available across separate commands, follow-up turns, and working-directory changes. Browser failures are surfaced instead of being hidden behind workarounds.
• More work can move at once. Paid-intro and subscription Organizations now share the same straightforward admission policy for available credits, so one active turn no longer needs to hold the whole balance. Ampere also has more capacity for independent AI Tool runs, helping ordinary bursts make progress with less unnecessary queueing. Temporary reservations remain distinct from credits already spent.
• Stopped work gets a more truthful receipt. When you stop a turn, Ampere reconciles the provider work already completed and records that amount once. If the evidence is incomplete, the receipt stays pending or unavailable instead of claiming that completed work cost nothing.
• DeepSeek Flash is current and steadier. When enabled for your Organization, the model picker now shows DeepSeek V4.1 Flash. Successful Flash responses no longer end as connection-closed errors solely because DeepSeek returned a newer model name, and the selected model remains the basis for a consistent credit receipt.

Why it matters

Less of your time goes to watching fixed waits, reconnecting tools, or finding work that Ampere already saved. You can start concurrent work with a clearer view of reserved and spent credits, return to interrupted sessions with more confidence, and review the result where you expected it.

Where to find it

Open Ampere to start a session, switch models, use the browser, or ask for a saved Project Asset. New Organizations will see the refreshed Brand Context studio during onboarding. Open Workspace → Usage to review your shared balance and recorded credit activity.