Skip to content

Phở Founder Lifetime: $59 once, 1 year of Phở AI included. Limited seats. See the offer

What's new

What is new in Phở, newest first. Only things you can use today.

Phở Fall ’26: the whole season on one page
  1. v2.18

    Systematic reviews to the PRISMA 2020 standard

    New

    • Keyboard screening (J/K to move, I include, E exclude, U undecided), 7 PICO-ordered exclusion reasons and a progress bar.
    • A PRISMA 2020 diagram with labelled stages and a button that copies a Methods paragraph from your own screening.
    • A characteristics-of-included-studies table you can sort by any column and download as CSV.
    • RIS export for EndNote, Zotero and Mendeley, plus CSV for Excel.
    • Scientific figures sized for 8 journal styles (Nature, IEEE, Elsevier, PLOS ONE and more), exported as PNG, PDF or TIFF.
    • Claude Fable 5.2 is in the model list, on your own API key on every plan.

    Improved

    • Discovery: sort by year or citations, filter by source and open access, a papers-per-year chart, abstracts inline.
    • Deep Research reports: two-way [n] citations, sources that follow the section you read, per-section rewrite and shorten, searchable references.
    • BibTeX has unique citation keys and correctly joined authors.
    • Reloading the page keeps your research progress and the open report.

    Fixed

    • Deep Research no longer hangs on “Starting…”, and Retry really retries.
    • Long reports no longer run the browser tab out of memory.
    • Runs on Claude 5.x models no longer fail at the start.
  2. v2.17

    Phở for sellers

    New

    • Connect your Shopify store and ask in plain words: sales, best-sellers, low stock. Every answer names the store and the date range.
    • Search products across every Shopify store to compare prices with competitors.

    Improved

    • Founder Lifetime now includes 1 year of Phở AI instead of 90 days.
    • Card and VietQR checkout are open to every visitor, no sign-in first.
    • Signing in to Research Tools now creates your Phở account too.

    Fixed

    • Business questions are no longer blocked by the safety check as off-topic.
    • Clinical mode is off by default and no longer offers a PubMed session on business questions.
    See it in Phở Fall ’26 →
  3. v2.16

    Phở wherever you work

    New

    • Open a dataset from Zenodo, Figshare, Dryad, Hugging Face or OpenML with its licence, schema, preview rows and citation.
    • A public API to call Phở’s research agent from your own code, with your existing Phở API key.
    • After paying by card you land on a welcome page that shows what to do first.

    Fixed

    • Research reviews cite real authors and years, in the language you asked for.
    See it in Phở Fall ’26 →
  4. v2.15

    Phở Subscription and a new logo

    New

    • Phở Subscription: Phở’s own AI with no key to manage. Pro $20, Max 5x $100 and Ultra $200 a month.
    • Run out mid-month and buy a credit pack instead of upgrading.
    • Ask for a small tool and get a working mini-app next to the chat.
    • Phở’s new star anise logo, everywhere in the app.

    Improved

    • Settings has Billing and Usage tabs, with plan changes right there.
    • A discount code field at checkout.
  5. v2.14

    Artifacts that stick around

    New

    • Artifacts now survive a page reload and stream straight into the side panel.
    • Lists show a quick preview of candidates and tell you when the list is ready.

    Improved

    • Attachments are read in parallel, so answers with files start sooner.

    Fixed

    • Real DOIs stay in the source list, and a tool step never ends without an answer.
  6. v2.13

    Research in one place

    New

    • One “Research” entry with two modes.
    • AI screening of the papers you already found.
    • 10 more interface languages, including Japanese, Korean and Chinese.
    • Choose how hard the model thinks, right next to the model name.

    Improved

    • A model manager in Settings and a shorter model picker.
    • Web search has its own Settings group, and Settings has a search box.
    • Deep Research reports are easier to read and in your language.

    Fixed

    • The reference checker no longer flags real papers as fake.
    • Streamed answers no longer end with runaway blank space.
  7. v2.12

    A new Settings

    New

    • Settings opens as a window with grouped navigation.
    • Spanish and Brazilian Portuguese interfaces.

    Improved

    • Answers come back in Spanish or Portuguese too, not just the menus.

    Fixed

    • Settings shows when your monthly budget resets.
    • Word (.docx) files are read, and PDF text is kept for follow-up questions.
  8. v2.11

    New

    • A searchable connector gallery with a page for every app.

    Improved

    • More workspace apps for research, notes and meetings.
  9. v2.10

    Bring your ChatGPT history

    New

    • Import your ChatGPT history: it is read in your browser.

    Fixed

    • An expired API key now says “key rejected” instead of a vague error.
  10. v2.9

    New

    • Message our support team from inside the app, or type /support.

    Fixed

    • Fixes for duplicated answers and a locked model in chat.
  11. v2.8

    Improved

    • Phở answers in your language, not the language of the prompt template.

    Fixed

    • Thinking notes no longer leak into saved answers.
    • Removed the confusing Temperature and Max tokens sliders.
  12. v2.7

    New

    • An answer that finished while you were away is picked up when you return.

    Fixed

    • Answers are recovered if your connection drops mid-reply.
  13. v2.6

    A denser sidebar

    New

    • A denser sidebar with a light/dark switch in the account menu.

    Improved

    • The model picker tells you which model will actually run.
  14. v2.5

    New

    • A Stop button for Deep Research.

    Improved

    • A run that runs out of time still gives you a report.
    • Finished reports are kept even if you close the browser.
  15. v2.4

    New

    • A warning when a cited paper has been retracted.
    • See which connected sources an answer used.
  16. v2.3

    A fresh chat look

    New

    • Search results appear source by source, with live progress.

    Improved

    • A cleaner message area: avatars, bubbles and spacing.
    • The Connectors page is now in Vietnamese.

    Fixed

    • One failing source no longer wipes out all search results.
  17. v2.2

    Improved

    • Smoother streaming answers.

    Fixed

    • Scanned PDFs can be read.
  18. v2.1

    New

    • Choose Light or Dark in Settings and see it change instantly.

    Improved

    • A clearly visible keyboard focus ring.

    Fixed

    • Fixed empty frames and stalled autoscroll while streaming.

  19. New

    • Deep Research adds ClinicalTrials.gov & EuropePMC. Autonomous research reports now also draw on the clinical-trials registry and the open biomedical full-text archive: every conclusion links back to a real source you can check.

  20. New

    • Clinical Answer Mode. One toggle in the composer: clinical questions are answered on real literature, with PubMed/DOI citations and GRADE evidence levels: and an honest label when no source is found. Free plan includes 5 grounded answers a day.
    • Source citations in chat, on every model. Turn on web search with any model: answers carry [n] markers and a source rail right beside them, so you can inspect every claim. Sources stay next to the text they support.
    • Bring your own key (BYOK). Connect your own OpenAI, Anthropic, Google, DeepSeek or xAI key: chat runs on your key at provider-direct pricing, with zero markup.
    • Editable canvas. Documents, data tables and artifacts can now be edited right in the canvas: change them by hand or ask AI to revise, updated in place.
    • Phở remembers your context. Knowledge you save to Brain is automatically recalled at the right moment in chat, with a transparent "Brain used" block showing exactly which notes informed the answer.
    • arXiv AI-bio feed. A rolling feed of the newest arXiv papers at the AI/ML × quantitative-biology intersection, filterable by topic. Fresh research beyond just ICML.
    • ICML 2026 collection. A venue landing page: intro, topics, and the spotlight papers: a clean entry point to all the AI-for-biomedicine work at ICML 2026.
    • Ask Phở on any document. Select any passage, or click one button, to ask Phở about a paper: it carries the whole context (title, authors, abstract) into the chat.

  21. New

    • Deeper paper reading. Each spotlight gets a Phở quick summary, key takeaways, and method/task tags. "Related" suggestions are smarter too (by shared tags and keywords).
    • Table of contents for long reads. Blog posts and case studies now have a table of contents for quick jumps, plus an "Ask Phở about this" button when you select text.
    • AI-bio Papers Radar. 315 curated AI-for-biomedicine papers from ICML 2026, filterable by 10 topics, each linking straight to the original and its code. Abstracts stay in the authors’ words.
    • Bilingual Vietnamese / English. All of Phở Labs is available in Vietnamese and English, with a language toggle on every page.
    • Phở Labs launch. labs.pho.chat is live: where Phở reads AI-bio research for you: a papers radar, a blog, and evidence-grounded clinical case studies.

  22. New

    • The 7-phase Research workspace. From literature search to screening, PRISMA, quantitative analysis and per-domain GRADE: in one workspace. Extracted numbers are confirmed by you before anything is computed.