Alfrada OS Docs

Appearance

Release Notes ​

User-facing product updates, new docs coverage, and notable workflow improvements.

How To Read These Notes ​

Each note answers four questions — in this order:

  • What you can do — concrete capabilities with specifics. Names, numbers, limits.
  • Where this shows up — real moments you'd actually hit this, in your voice.
  • Try it — copy-ready prompts you can paste today.
  • Heads up — only when there's something you genuinely need to know: pricing, limits, breaking behaviour, migration.

Current Releases ​

October 19, 2026 ​

  • DeepSeek V4.1 Flash Moves To The October 17 Snapshot — DeepSeek V4.1 Flash now runs on DeepSeek's October 17 snapshot: same picker row, same 1M context, Budget 1×, and zero-retention hosting, with a sharper model underneath (Artificial Analysis intelligence index 39.5, up from 24.7). Chats already pinned to V4.1 Flash, and older MiniMax M3 and Nemotron 3 Ultra pins, roll forward automatically.

October 16, 2026 ​

  • New Decks And Word Documents Build In The Dashboards Panel — Editable PPTX And DOCX — New presentations build as 1280×720 slides and new Word files as A4 or Letter pages, both live in the Dashboards panel. Download them as editable PowerPoint or Word: real text, headings, lists and tables, with charts as pictures. The download link comes in the same reply. Asking for a change re-exports the same file, and a saved template can lend its colours, fonts and logo to a fresh layout. Uploaded decks and Word templates still go through PPTX Patch and Doc Patch.
  • Claude Haiku 5.5 and Nemotron 3.5 Super VL — Claude Haiku 5.5 is in Fast on Amazon Bedrock: 1M context, EU geo profile, zero data retention, Budget 1×. Nemotron 3.5 Super VL is the new vision-language Super on OpenRouter, also Budget 1× and pinned to Zero Data Retention. Nemotron 3 Ultra and Nemotron 3 Super 120B are retired: their chats carry on in DeepSeek V4.1 Flash and Claude Haiku 5.5. The zero-retention Claude models — Opus 5.5, Sonnet 5.5, Sonnet 5, Sonnet 4.6, and Haiku 5.5 — show the blue EU flag, and Sonnet 5.5 is no longer hidden when EU Data Residency is on.

October 15, 2026 ​

  • Unopened Swarm Results Now Come To You By Email — If you haven't opened a finished Swarm run within 5 minutes, your agent emails you the deliverables and an executive summary. Local Gemma picks the files that matter, at no token cost. Not sent for scheduled tasks or EU Data Residency accounts, and you can turn it off in Safety Center.
  • Swarm Plans Start On Their Own After 2 Minutes — An unanswered Swarm plan review now counts as approved after 2 minutes, so a fire-and-forget task runs even with the tab closed. A countdown shows the time left and Cancel still stops it. Mid-run re-approvals and planner questions get the same clock, and after 10 automatic approvals in a row the next review waits for you.

October 14, 2026 ​

October 11, 2026 ​

  • One-Time Top-Up Packs From $10 — Up To 50% More Tokens — Four one-time packs, no subscription: $10 → 12M, $25 → 35M (+17%), $50 → 80M (+33%), $100 → 180M (+50%). Buy as many as you want, on every plan including Community. Pack tokens last 6 months from your latest top-up, and each pack shows what it typically covers. When you run low, the chat banner shows the top-up and the next plan up side by side, and on Community any pack unlocks premium models. The $5 pack and the $10 and $20 monthly packs are retired; what you already bought keeps working.

October 9, 2026 ​

  • Speech and Dialogue Move to Eleven v4 — Same Media Band — Speech, chat voice-overs, and Dialogue now generate on Eleven v4. Alfred, Ada, and the six delivery styles are unchanged. If v4 cannot take a Speech call, Alfrada tries Eleven v4 Turbo, then the older multilingual model. You do not pick the model. Still Pro and above, 500,000 tokens per call.

  • Dialogue — Two Voices, One Recording — Voice a script in one call on Eleven v4. Alfred and Ada use their own voices; anyone else uses an ElevenLabs voice id you already have. Set the language, including Hindi. Long scripts are joined into one MP3 with a start and end for each line. Put each person on screen for those times and mix that MP3 under the picture as the voice. Pro and above, 500,000 tokens per call, the same as Speech.

  • Code Executor — Twice The Compute, A Headless Browser, And More Libraries — Python sandboxes now run on 4 vCPUs, 8 GB RAM, and 10 GB disk (double the CPU and memory), numpy/scipy maths use one thread per core (a large matrix multiply went from 1.1 s to 0.16–0.25 s), and sandboxes start from a pre-built environment, so a new conversation's first run no longer waits on a rebuild. Playwright with headless Chromium is pre-installed for rendering HTML to PNG/PDF and reading JavaScript-heavy pages, alongside polars, DuckDB, statsmodels, sympy, networkx, pypdf, pdfplumber, reportlab, olefile, requests, httpx, and lxml. Price per run is unchanged.

October 5, 2026 ​

  • Music Lab Moves to Suno v6 — Three Models, Same Studio Band — Music Lab now generates on Suno's v6 family. V6 is the default; ask for something experimental and you get V6_WILD, ask for it fast and you get V6_MINI. Idea mode, Structured mode, vocal gender, instrumental-only, and two variants per call are unchanged. V3_5, V4_5 and V5 are retired with nothing to migrate. Still Pro and above in the unchanged Studio band — but idea-mode prompts now cap at 3K characters (Structured mode still takes 5K).

October 1, 2026 ​

  • GPT-6.1 Sol Replaces GPT-6 Sol — Same Premium Band — GPT-6.1 Sol takes GPT-6 Sol's Recommended seat at the unchanged Premium 6× band: close to GPT-6 Astra on agentic coding, computer use, and professional work, and a clear step up from GPT-6 Sol on programming, debugging, and document work. Same 1.05M context, 128K output, tools, Swarm, Auto eligibility, and complexity-aware reasoning. GPT-6 Sol is retired — pinned chats and saved preferences roll forward automatically. GPT-6 Luna, GPT-6 Astra, and GPT-5.6 Terra are unchanged.

  • Presenter Mode — Anyone On Camera, Over Your Screen — Put a person in the corner of a screen recording and ship the file. Generate Alfred, Ada, or anyone you register (from a few photos or a short clip), switch presenter between lines with the Script lane, move between full screen and a corner box, freeze the screen and draw on the held frame. Pro and above.

September 29, 2026 ​

  • Claude Sonnet 5.5 — Faster Sonnet, Same Premium Band — Claude Sonnet 5.5 joins the picker next to Sonnet 5 at the unchanged Premium 6× band: a clear step up on coding and knowledge work, 30%+ faster output, fewer tool calls per task, 1M context, 128K output, Workhorse Auto routing, and Swarm eligibility. Global Bedrock routing only for now, so it is hidden under EU Data Residency; Sonnet 5 is not retired and stays live for EU-only accounts.

  • Jira — Triage, File, and Move Tickets From Chat — Jira Cloud connects from Integrations: search, file, comment, attach, transition, and run boards and sprints from the conversation. Writes ask before they run. On the desktop app, Jira Server and Data Center behind a VPN use a Vault login, including a bulk export of a queue to spreadsheet files.

September 23, 2026 ​

  • GPT-6 Sol and Luna Replace GPT-5.6 Sol and Luna — Sol Moves to Premium 6× — GPT-6 Sol takes GPT-5.6 Sol's Recommended seat at Premium 6× (down from SOTA 12×) and GPT-6 Luna takes 5.6 Luna's place under Fast at Budget 1×: same 1.05M context, 128K output, tools, Swarm, and complexity-aware reasoning. Sol stays Auto-eligible; Luna is a pinned pick and Swarm worker. GPT-5.6 Sol and Luna are retired — pinned chats and Argus heartbeats roll forward automatically. GPT-5.6 Terra is unchanged.

September 22, 2026 ​

  • Claude Opus 5.5 Replaces Opus 5 — Same SOTA Band — Claude Opus 5.5 takes Opus 5's seat on Recommended and in Closed SOTA, still at SOTA 12×: adaptive thinking, 1M context, 128K output, Swarm and Auto eligible, EU-only geofencing supported. Opus 5 is retired. Pinned Opus 5 and Opus 4.8 conversations roll forward to 5.5 automatically.

  • Grok 4.7 — Takes the Recommended Seat from 4.6 — xAI's Grok 4.7 replaces Grok 4.6 on the Recommended strip: same 500K context, image and file input, always-on reasoning, full tool support, and every call still pinned to Zero Data Retention endpoints. Built for long-running software engineering and for verifying its own work. Auto-routed at the SOTA tier, Swarm-eligible. Grok 4.6 is retired — existing sessions pinned to it keep working.

September 19, 2026 ​

  • Transport Modelling — Calibrated Base Models, Place Names, Monte Carlo — Name a place and Alfrada builds the world model: base demand from population and cited trip rates for every mode, rule-based signals, calibration to observed journey times or your counts (GEH), and a TAG / TfL / FHWA-structured report. New: Monte Carlo over distributions built from your own words (P10 / P50 / P90, probability the scheme is worse), parallel queued simulations, a study-area outline and hide-interface button in the 3D replay, and an Experimental tag on every call.

September 15, 2026 ​

  • DeepSeek V4.1 Flash — On Recommended, Budget 1× — DeepSeek V4.1 Flash joins Fast and Recommended at Budget 1×: 1M context, Swarm-eligible, Auto-routed at quality 60, pinned to zero-retention OpenRouter hosts that support tools. It takes Gemini 3.8 Flash's Recommended seat (Gemini stays under Fast). Claude Opus 5 is visible on that strip.

  • PDF Creator — Every Page You Can Build for the Web, Now as a PDF — PDFs are now authored as a web page and printed with headless Chromium, replacing the hand-coded builder behind every PDF until today. Real CSS layouts, any Google Font embedded, the full Unicode range (no more ✓ turning into [x]), live charts at print resolution. You pick the page size (A4 through Ledger, or custom), orientation, margins, and running headers/footers with live page numbers — and slide_16x9 renders a genuine 16:9 deck. New Word documents move to Word Ops.

September 14, 2026 ​

  • Now Alfrada Can Search, Analyse and Recommend Mobile Apps — Two tools cover the whole phone market. Apple App Store searches Apple's store for iPhone, iPad and Mac apps and returns rating, review count, age rating, size and version history. Google Play Store searches Android apps and games, reads the top free / paid / grossing charts, opens any listing in full (download tier, review histogram, permissions, data safety) and reads its user reviews newest-first or by star rating. Both quote prices in your own storefront's currency by default, or any market you name.

September 10, 2026 ​

September 9, 2026 ​

  • Swarm Can Now Seat A Panel — Not Just Split The Work — Ask for a roundtable, a focus group, or "what would each of these people say," and Swarm seats 2 to 5 named personas with a moderator. Debate is still the binary tool; a panel is for when the call is not a coin flip.

September 7, 2026 ​

  • AutoScout24 — Search Europe's Biggest Car Market, and Know If the Price Is Fair — Search used and new cars across eight European markets (de, at, be, es, fr, it, lu, nl) with filters for make, price, mileage, registration year, power, fuel, gearbox, body type, seller and radius — then pull any listing's full detail sheet with median_price, AutoScout24's own valuation, and a plain above/below-market verdict.

September 5, 2026 ​

  • GPT-6 Astra Is Live — OpenAI's Flagship At Apex — GPT-6 Astra joins Closed SOTA: 1.05M context, 128K output, quality 71, Apex 18×. Independent boards already put it near the top — ARC Prize SOTA on ARC-AGI-2/3, Artificial Analysis Intelligence Index #2. Note (September 9): Auto and Swarm no longer select it; pin it yourself.

September 3, 2026 ​

  • Gemini 3.8 Flash — Longer-Horizon Coding At Flash Cost — Google's newest Flash workhorse brings 1M context, 65K output, multimodal input, tools, structured output, caching, and Swarm worker support at Workhorse 3×. Gemini 3.7 is retired; pinned 3.5, 3.6, and 3.7 sessions roll forward to 3.8 automatically.

September 2, 2026 ​

  • GLM 5.3 Answers Sooner — High Effort, Flash On Low — GLM 5.3 now thinks at high effort and GLM 5.3 Flash at low. Beast — pinned or Auto-selected — steps them to max / high. Thinking stays on (Z.ai does not allow disabling it). First tokens arrive sooner than the previous max default.
  • Claude Fable 5.1 Replaces Fable 5 — Same Slot, Same Pricing — Claude Fable 5.1 takes Fable 5's picker slot: Mythos-class, 1M context, 128K output, adaptive thinking, same pricing, still a deliberate non-ZDR manual pick (kept out of Auto). Pinned Fable 5 sessions roll forward automatically.
  • Community Starts On Budget 1× Models — New Community accounts start on Budget 1× only. Auto, Swarm, and scheduled runs stay on that band until review approval or a purchase. Paid plans are unchanged.

September 1, 2026 ​

  • OneDrive — Find, Move, Share, And Version Files From Chat — Connect up to three Microsoft accounts; search and browse OneDrive, upload or download session files, organize folders, create controlled sharing links, revoke permissions, restore earlier file versions, and recover deleted items in OneDrive Personal.

August 30, 2026 ​

August 27, 2026 ​

  • Workbooks And Word Files Now Fill Like The File You Sent — A saved Excel or Word template is filled in place: inspect seeds a working copy, are replaced, leftover placeholders are reported, and a visual check flags clipped columns and overflow. Numeric strings and formula text stay live (no #VALUE! from a missing =). Presets still build a new workbook when no template matches. Auto-fit columns and session images land in the sheet; Word create can embed a logo.

August 26, 2026 ​

  • GLM 5.3 Flash Is Live — Native Multimodal, Budget Band — GLM 5.3 Flash is on the Fast tab via Z.ai's own API: 1M context, image input, tools, always-on low-effort reasoning, Budget 1×. Community Auto can pick it; paid Auto skips it (quality 58, under the paid floor of 60). Pin it when you want the GLM family without Premium 6×. GLM 5.3 stays the flagship.

  • Paid Auto Stays On Capable Models — On Pro, Pro Max, and Black, Auto never uses a budget model. It picks the cheapest healthy model at quality 60 or above. Community Auto still keeps the cheap default for everyday asks.

  • Uploaded Decks And PDFs Now Patch Like The File You Sent — PPTX Patch now updates native charts, table cells, template hero photos, fills, fonts, and slide order, then visually checks changed slides. PDF Patch fills forms, merges and splits pages, stamps and watermarks, and restamps small text defects in place. Playbooks and the tool library list the real surface — not the June-era "swap a logo" subset.

August 19, 2026 ​

  • GLM 5.3 Is Live — Direct Z.ai, Neck And Neck With Fable — GLM 5.3 replaces GLM 5.2 as Z.ai's flagship, called on Z.ai's own API: 1M context, always-on high-effort reasoning, 128K output (up from 65K), tools, Swarm, and SOTA Auto at the same Premium 6× band. Auto quality is scored even with Claude Fable 5. On Can it Run Alfrada? it debuts #5 (8.55 composite). Pinned 5.2 sessions roll forward automatically. Updated Aug 20: routing moved from OpenRouter to the direct Z.ai API.

  • Long Renders Keep Going — The Chat Does Not Freeze — Video Lab, Music Lab, Image Lab batches of 2+ images, heavy Video Editor renders, and long Code Executor runs detach into a live card in this thread. The conversation stays usable; reload shows running / delivering / finished / failed / cancelled instead of a forever spinner. List or cancel with background_job. Up to 10 distinct Image Lab prompts render as one background batch.

  • Video Editor — Reframe, Jump-Cut, Captions, Overlays, And Animated Slides — One request covers reframe to 9:16, silence jump-cuts, karaoke or switchable captions, timed lower-thirds, soundtrack extract, scene split, Manim explainer slides, and Ken Burns stills. Ingest now transcribes and labels speakers. Long renders hand off to a background job. Finish the cut in Live → Video Studio.

  • Image Lab Defaults To Seedream — Gemini For 4K And Batches — A normal Image Lab call now renders on ByteDance Seedream 5.0 Pro. Ask for 4K or several images and auto switches to Gemini. A refusal on the default route retries once on Grok Imagine. Pin seedream / gemini / xai when you want a specific path. Higgsfield Soul is no longer the automatic fallback.

  • History Search Now Finds Files, Tool Results, And Other Conversations — Mixed search returns conversation, uploads, generated files, and past tool results together. An empty new chat no longer dies as a silent miss: it shows the top 5 hits from other sessions, clearly labelled. Fuzzy filename lookup across all sessions; * is for list/find/transcript paging, not a semantic query.

August 18, 2026 ​

  • Google Jobs — Every Board's Openings, With the Pay Attached — New google_jobs tool returns live listings Google pooled from LinkedIn, Indeed, company career pages and job boards, each with salary, schedule, posting age, remote and benefit flags, the qualification bullets Google extracted, and a direct apply link. Filters go in the sentence ("remote", "part time", "since yesterday"), up to 30 listings per search, and every set saves as a CSV carrying the complete description text.

  • Google Trends — Ask What the World Is Actually Searching For — New google_trends tool returns Google's 0–100 relative interest index: interest over time (with peak date and a rising/flat/falling read), interest by country, state, metro or city, and top + rising related queries and topics. Compare up to 5 terms on one scale, measure web, images, news, shopping or YouTube, pick any of 9 time ranges or a custom window, and get every result saved as a CSV in the session for charting and reports.

  • Rank Tracking — Where a Site Sits on Google, By Keyword — New google_rank_tracking tool returns the live organic position for a named domain (and competitors) on up to 10 keywords, scanning the top 100, for a country/city and desktop/mobile/tablet. Omit the domain to see who owns page one. Every run saves a CSV of every scanned row for before/after tracking.

August 17, 2026 ​

August 16, 2026 ​

  • Hours Saved Now Reads On Gemma 4 12B — The nightly Hours Saved analyst moves from Gemma 4 e4b to Gemma 4 12B on local Ollama. New conversation-days get a stronger profession / hours / satisfaction read; transcripts still never leave the box. Already-closed e4b receipts stay as written.

August 15, 2026 ​

  • Claude Sonnet 5 Works With EU Data Residency — Claude Sonnet 5 - Reasoning is now available to accounts running EU Data Residency. AWS published an EU inference profile after the model launched, so EU requests are geofenced to eu-west-2 rather than refused. Sonnet 5 returns to the picker, to Workhorse Auto routing, and to Swarm worker steps for EU-only accounts, at the unchanged Premium 6× band. EU-only accounts also now see the EU-hosted Qwen3.8 Max and Qwen3.7 Flash in the picker's Recommended list. Supersedes the Global-only note from July 1.
  • Browser Takeover — Jump In When the Web Agent Gets Stuck — When the web agent hits something only a human can clear — a CAPTCHA the automatic solver couldn't pass, a passkey or "approve on your phone" prompt, a QR sign-in, a login with no saved credentials — it now hands you the live browser in a "Browser Needs You" window. Clear the step, click Done, and the agent continues on the same page; nothing you type there is visible to the assistant. The browser also stays open between calls in a conversation (10 min idle / 60 min max), so log-in → read-the-code → enter-the-code flows work end to end, and typed-code prompts now wait 3 minutes.

August 14, 2026 ​

  • Image Consistency — Same Person, Same Object, Same Style, Verified Locally — New local tool compares 2–4 session images in one pass and returns a structured same-person / same-object / style-drift verdict with confidence and specific differences, powered by Gemma 4 12B on local Ollama (images never leave the box, no per-call cost). The video editor's restyle pipeline now self-audits: every restyle attaches an advisory frame-consistency verdict to the result.
  • Gemini 3.7 Flash — Smarter Workhorse, Same Band — Gemini 3.7 Flash replaces 3.6 Flash as the public Workhorse Flash (3×). Pinned 3.6 and 3.5 sessions roll forward automatically.

August 13, 2026 ​

  • Grok 4.6 — xAI's Frontier Reasoner Joins the Picker at a Workhorse Band — Grok 4.6 replaces the hidden Grok 4.5 in the Closed-Weight SOTA section: 500K context, image input, always-on reasoning, tools, Swarm and SOTA-tier Auto eligible, ZDR-pinned through OpenRouter — and billed in the Workhorse 3× band. It ranks #3 on Can it Run Alfrada? (8.63 composite), behind only Opus 5 and GPT-5.6 Sol. Global route only; Grok 4.5 sessions keep resolving but should re-pin.
  • Qwen3.8 2.4T A95B — Open Weights With Enforced ZDR Routing — The text-only open-weight Qwen3.8 model joins the picker through OpenRouter with 1M context, mandatory reasoning, tools, structured outputs, Swarm eligibility, and a Workhorse 3× token band. Alfrada OS requires OpenRouter to use a Zero Data Retention endpoint for every request; this global route is manually selected and remains separate from the multimodal, Auto-eligible, direct EU Qwen 3.8 Max service.

August 11, 2026 ​

  • The Agent Can Suggest Swarm — You Decide — When a turn turns out to need parallel workers, the agent can suggest moving it to Swarm mid-task — and a consent banner pauses the turn until you decide, on every session type. A new Safety Center setting (Ask / Always allow / Never) skips the question in either direction. Once escalated the session stays in Swarm, with short follow-ups dropping back to the single agent on their own.

August 9, 2026 ​

August 4, 2026 ​

  • Qwen 3.8 Max + Qwen 3.7 Flash — Frontier Depth Or Multimodal Speed — Two direct EU Qwen models join the picker with 1M context, 128K maximum output, multimodal input, reasoning, tools, and structured output: Qwen 3.8 Max at Premium 6× (SOTA Auto + Swarm), and Qwen 3.7 Flash at Light 2× (budget Auto). Qwen 3.6 Flash and Qwen 3.7 Max retire from the picker with pinned sessions rolling forward.

August 1, 2026 ​

  • DeepSeek V4 Flash 0731 + GPT-5.6 Luna & Terra Price Cuts — DeepSeek's official V4 Flash 0731 build lands in the picker at the #3 recommended slot (same 284B / 13B MoE, 1M context, budget tier). We've cut GPT-5.6 token rates: Luna moves Workhorse 3× → Budget 1×, Terra moves Premium 6× → Workhorse 3×; Sol stays SOTA 12×.

July 25, 2026 ​

  • Claude Opus 5 Is Live — Anthropic's New Frontier Flagship — Claude Opus 5 joins the picker at the top of the SOTA tier (12×): near-Fable-5 intelligence one band below Fable's, adaptive thinking with visible summaries, 1M-token context, 128K output, Swarm and Auto eligible, with EU-only geofencing supported. It replaces Opus 4.8 in the same token band — Opus 4.8 is retired from the picker and pinned conversations roll forward automatically. For harness scores and evidence packages, see Can it Run Alfrada?.

July 22, 2026 ​

July 20, 2026 ​

  • Safer Frontier-Model Access For New Community Accounts — New Community accounts can start with every model up to Workhorse 3×; higher-cost bands unlock after an automated, locally run Gemma 4 e4b account review and operator approval, or immediately after a verified purchase. Existing accounts and paid plans keep their access.
  • Can It Run Alfrada? — Inspect The Work, Not Just The Score — The public LLM Benchmark compares nine models completing two production knowledge-work jobs, with 126 file-by-file judge verdicts, evidence packages, calibrated rankings, cost, and speed. Inkling is now manually selectable, but remains outside Auto routing and Swarm after an uneven 5.60 result exposed material arithmetic, consistency, and reproducibility failures.

July 17, 2026 ​

  • Kimi K3 Is Live — Moonshot's Flagship Multimodal Reasoner — Kimi K3 joins the picker as Moonshot's 2.8T-parameter open-weight flagship: 1M-token context, 50K output tokens, always-on thinking at maximum effort, text + image input, tool calling, Swarm eligibility, Auto routing, and billing at the Premium band (6×) — not 12×. It takes the #2 recommended slot after GLM 5.2; the recommended row grows to four picks and Kimi K2.7 Code leaves it but stays selectable.

July 16, 2026 ​

  • Daytona Can Now Operate Your Servers Without Seeing The Keys — Save server access in the Vault and the existing Code Execution capability can pull logs, inspect health, or deploy through a dedicated ephemeral Daytona sandbox. Secrets enter only its fixed runner process; reads run immediately, writes pause for approval, and longer work can set a 5-minute–24-hour idle window plus a wake_me follow-up.

July 15, 2026 ​

  • Smarter Auto Picks + Updated Qwen Token Rates — Auto now selects the lowest-cost healthy model that clears your task's quality requirement and holds that level as work gets harder. Qwen 3.7 Max is now Premium 6×, Qwen 3.6 Flash is Light 2×, and Claude Fable 5 drops to Apex 18×.
  • Argus Heartbeat Default — GPT-5.6 Luna — Argus scheduled check-ins now default to GPT-5.6 Luna instead of GPT-5.4 Mini; existing Mini-default heartbeats migrate automatically, and you can still pick Auto or any other model.

July 14, 2026 ​

  • Value Modelling — See The Expert Hours & Dollars Alfrada OS Saves You — Settings → Account → Hours Saved shows expert hours saved, dollar value at per-profession rates you can edit, compression multiples, and per-conversation receipts with satisfaction labels. Analysis now runs on local Gemma 4 12B (private, tool I/O stripped; upgraded August 16), follows Anthropic's Economic Index methodology (task-level value at expert wage rates, conservatively flexed 50%), and backfills up to 90 days on demand.

July 13, 2026 ​

July 10, 2026 ​

  • GPT-5.6 Is Live — Luna, Terra and Sol Scale to the Job — three GPT-5.6 models join the picker with 1.05M context, 128K output, tools, Swarm, and complexity-aware reasoning: Luna for fast high-volume work (Budget 1×), Terra for everyday coding and agents (Premium 6×), and Sol for the hardest reasoning and multi-step workflows (SOTA 12×). Auto can route across all three; Luna can also work inside multi-agent runs.

July 6, 2026 ​

  • GLM 5.2 Now Routes Direct To Z.ai — Free This Weekend — GLM 5.2 now uses a direct Z.ai API route (OpenRouter vendor retired); same 1M context, reasoning, tools, and swarm eligibility with lower latency. Free Fri–Sun 10–12 July 2026 (UTC) — zero token-budget and usage cost — then automatic return to 6× premium billing from 13 July (UTC). Pick GLM 5.2 in the selector; the old GLM 5.2 (via OpenRouter) entry is gone.

July 5, 2026 ​

  • Memory Now Remembers How Things Changed — long-term memory becomes a version-controlled timeline: every change is a dated commit with a before/after diff (+ created / ~ superseded / − invalidated / ± edited) attributed to you, the agent, a migration, or a re-mine. Facts are superseded, not overwritten — the old one is kept and stamped "true until…" while the new one takes over. New Timeline and Entities views (a lane per person/org/project/place) live in Digital Twin → Memories, with a per-fact history drawer. A background, billing-exempt, resumable re-mine rebuilds your timeline from your whole chat history, dating each fact to the conversation it came from. Existing memories import automatically as dated genesis commits; superseded ≠ deleted.

July 4, 2026 ​

  • Wake Me — Alfrada OS Checks Back In The Same Conversation — wake_me schedules a wake-up that resumes this same chat after 5 minutes to 24 hours so Alfrada OS can follow through on time-dependent work (awaited email, long render, intraday check) without polling or opening a new session. One pending wake per conversation; 10 consecutive self-wake turns max; five pending wakes per user. Always on and free. Not WhatsApp/Argus — use Task Scheduler for recurring or clock-anchored jobs.

July 1, 2026 ​

  • Claude Fable 5 (Mythos) Is Back — Anthropic's Strongest Model, Selectable Again — after the June 13 pullback, Claude Fable 5 (Mythos) returns as a deliberate manual pick [Not Private]: highest quality rating of any Alfrada OS model (above Opus 4.8), adaptive thinking, 1M context, 128K output, and the Overkill band (24× tokens). Still non-ZDR (provider may retain inputs/outputs) so it's kept out of Auto routing; routed through Bedrock's Mantle endpoint (Global by default), with EU-only geofencing supported. Supersedes the June 10 and June 13 notes.
  • Claude Sonnet 5 Is Live — Anthropic's New Workhorse for Coding and Agents — Claude Sonnet 5 joins the picker with 1M context, adaptive thinking, 128K output, Premium-band pricing (same rates as Sonnet 4.6), Workhorse Auto routing, and swarm eligibility. Originally Global Bedrock only — superseded by the August 15 note, which adds EU Data Residency support.

June 28, 2026 ​

  • Memory & Playbooks, Now Focused Tools — long-term memory's single multi-purpose tool is now five focused tools: memory_search (search or list), memory_create, memory_update, memory_delete, and a dedicated playbook tool. Saves now require the namespace and text together, so a memory either lands completely or says exactly what's missing — no more half-saved entries or retry loops, on any model. Same "Agent Intuition" feature; existing memories and playbooks carry over and default playbooks auto-update to the new names.
  • Self-Checking Answers + Pick How Hard Alfrada OS Works — on bigger tasks, Alfrada OS now writes a task-specific rubric, grades its own draft against it, and revises before you see it. Medium does one self-review pass; Beast keeps revising until the rubric clears and also runs a vision pass over generated charts/images (axes, legends, cut-off text). A new performance slider lets you pin Fast / Medium / Beast or leave it on Auto, with a badge showing the mode that actually ran and sticky persistence across the session. Trivial turns are never graded. Grader usage is billed.

June 17, 2026 ​

  • GLM 5.2 Is Live — Z.ai's Long-Horizon Agent Flagship — GLM 5.2 joins the picker as Z.ai's upgraded long-horizon agent flagship: 1M-token context (up from GLM 5.1's ~203K), reasoning enabled, tool support, swarm eligibility, and SOTA-band placement at the same premium pricing as GLM 5.1. Suited for project-level software engineering, long-running agent workflows, and complex multi-step automation. Auto-routable in the SOTA tier on every plan (scored on quality and speed), and directly selectable too.

June 13, 2026 ​

  • Kimi K2.7 Code Is Live, Fable 5 Pulled Back — Kimi K2.7 Code replaces Kimi K2.6 in the public specs sheet as Moonshot's coding-focused SOTA model: 262K context, reasoning enabled, tool support, and swarm eligibility. Anthropic Claude Fable 5 (Mythos) has been pulled back after a DoW notification and is no longer selectable while the notice and related obligations are reviewed.

June 10, 2026 ​

June 9, 2026 ​

  • Edit Uploaded Word Documents In Place, Formatting Intact — the new Doc Patch tool surgically edits a .docx you uploaded instead of rebuilding it: inspect every paragraph, table, image, and comment with stable references, then replace exact text while keeping run formatting, patch table cells and rows, swap an image, or insert/delete blocks — preserving surrounding text and table styles. It can also author a brand-new .docx from structured sections. Every edit runs against a required preservation contract and returns an honest change report. Use Report Generator for net-new reports from markdown. Free on every plan.
  • Edit Uploaded PowerPoint Decks In Place, Design Intact — the new PPTX Patch tool surgically edits a .pptx you uploaded instead of rebuilding it: inspect every slide's shapes, text, image positions, notes, and animation XML, then replace text, swap an image while keeping its exact size/crop, move/resize/hide/delete a shape, duplicate a slide, or update notes — preserving the original design, masters, and existing animations. Every patch runs against a required preservation contract and returns an honest report of what changed and what couldn't be guaranteed. Use Presentation for brand-new decks; new-animation authoring is not supported. Free on every plan.

June 6, 2026 ​

  • Nemotron 3 Ultra Is Live — NVIDIA's Frontier Orchestration Reasoner — Nemotron 3 Ultra [Global] lands in the picker: NVIDIA's 550B / 55B active hybrid Transformer-Mamba MoE for long-running agentic planning, 256K context, tool support, swarm eligibility, and Workhorse-band pricing. In the Auto routing pool on every plan; global via OpenRouter — use Nemotron 3 Super 120B [EU] when you need EU residency.

June 3, 2026 ​

  • Live Video Studio — Edit Chat-Built Videos Without Starting Over — the Live tab now opens Video Studio while Alfred runs video_editor or video_lab: one panel per session, a step list during the build, then preview + horizontal timeline (drag clips, trim, transitions, colour grade, audio, subs) after finalize. Update preview rebuilds your edits on the CUDA sidecar.

June 02, 2026 ​

  • MiniMax M3 Is Live — Long-Context Workhorse For Agentic Tasks — MiniMax M3 is now available as a Workhorse-tier model: 524K context, tool support, swarm eligibility, budget-band pricing, and a conservative quality score based on the first judged leaderboard run. Strong first signal on baseline tool orchestration; not promoted to SOTA yet because deep prediction and swarm timed out under the benchmark harness.

May 30, 2026 ​

  • HTML Dashboards — Beautiful, Interactive Reports That Render Live And Export To PDF — the new HTML Dashboard tool builds live, interactive dashboards (KPI walls, research readouts, comparisons, roadmaps) from six prebaked templates, renders them in a new Dashboards panel, refreshes the same dashboard in place when data changes, and exports to PDF / PNG / HTML. Self-contained (Chart.js, inline data), on-brand with auto dark mode, free on every plan.

May 29, 2026 ​

  • Claude Opus 4.8 Is Live — Anthropic's New Frontier Flagship — Claude Opus 4.8 is in the picker: Anthropic's most capable frontier model with adaptive thinking, a 1M-token context window, and the highest quality rating of any model Alfrada OS offers. Built for strategic planning, financial analysis, and deep research; swarm-capable; EU-only geofencing supported for in-region data residency. Replaces Opus 4.7 and 4.6 at the same price. Available on every package.

May 25, 2026 ​

May 23, 2026 ​

May 20, 2026 ​

  • Auto Just Got Cheaper — Gemini 3.1 Flash-Lite Is The New Default, Gemini 3.5 Flash Joins The Picker — Auto's default model swaps from gemini-3-flash-preview to gemini-3.1-flash-lite at a lower token rate, with the same 1M context and tool support. Background intent/Q&A/summarisation moved off Gemini 2.5 Flash-Lite onto 3.1 Flash-Lite in the same step. Separately, Gemini 3.5 Flash lands as a Workhorse-tier pick — 1M context, reasoning on at medium effort, swarm- and worker-eligible — for harder agentic loops where you want frontier Flash quality at a Workhorse band cost.
  • xAI Grok Imagine Joins Image Lab Through OpenRouter — pin provider: "xai" in Image Lab to render 1K/2K posters, ads, packaging, menus, social graphics, and reference-image edits through x-ai/grok-imagine-image-quality on OpenRouter. Gemini remains the default path, Higgsfield Soul remains the moderation fallback, and xAI's named-entity rendering is documented as capability rather than legal clearance.

May 19, 2026 ​

  • xAI Imagine Video Joins Video Lab — Native-Audio Clips With A Safety-Net Fallback Chain — Video Lab now talks to three providers (Runway, Higgsfield, xAI). Pin provider: "xai" for short native-audio clips on x-ai/grok-imagine-video via OpenRouter (1–15s, 480p/720p, eight aspect ratios, up to seven reference images). provider: "auto" now runs a three-stage fallback chain — xAI → Runway → Higgsfield — with an Image Lab bridge so a text-to-video request never silently falls back to a Google image search. Every auto response carries a fallback_chain block showing which provider succeeded.

May 18, 2026 ​

  • Image Lab Now Auto-Falls-Back To Higgsfield Soul When Gemini Blocks — when Gemini's safety filter refuses an image (named franchises, IP overlap, silent refusals), Alfrada OS now quietly re-runs the same prompt on Higgsfield Soul and tells you which model produced the file. Pre-emptive prompt rewriting catches known-blocking combos (real person + named IP + photoreal) before submission, and you can chain refinements ("darker lighting", "more rain") through Soul without ever going back to Gemini.
  • Higgsfield Joins Video Lab — Cinematic Camera Moves Runway Can't Quite Do — pin provider: "higgsfield" on any video_lab call for native cinematic camera arcs and landscape pans. Four Higgsfield models ship today: DOP Lite for previews, DOP Turbo for working drafts, DOP Standard for keepers, and Kling 2.1 Pro for landscape pans (5 or 10 seconds). The new video_provider_routing playbook explains when to pick which provider; failures surface with retry suggestions and never silently fall back to Runway.
  • Google News — Localised News Search That Cites Itself — google_news runs Google News via SearchAPI with country/language auto-localised to your detected location (Berlin → German edition, London → UK edition, Karachi → Pakistan edition). Top 5 articles auto-save as .news.json citations in the session out/ folder, indexed for report_generator and presentation. Thumbnails decode from SearchAPI's base64 data once on the backend so they render in the chat card without inflating LLM token cost. Filters: recency (last_hour through last_year), sort order (most_recent or relevance), pagination. Per-row inline citation surfaces the publisher hostname + saved filename next to each article.

May 16, 2026 ​

  • Google Images — Localised Image Search That Saves Straight Into Your Session — google_images runs Google Images via SearchAPI with country/language auto-localised to your detected location; top 5 results download into out/ by default and are indexed for presentation, video editor, and report generator. Filters cover image size (16 options including megapixel thresholds), color (15), image type (photo, clipart, line_drawing, gif, face), aspect ratio, recency, usage rights (Creative Commons / commercial), SafeSearch, and pagination. Each saved file ships with a credit line for inline rendering.
  • Amazon Search & Amazon Product — Localised Shopping On Your Storefront — amazon_search and amazon_product ship together; both auto-route to the user's local Amazon (amazon.de from Berlin, amazon.co.uk from London, amazon.in from Bangalore) across 21 storefronts. Results are filtered to items deliverable to the user's country. Search supports sort orders (featured / price / review / newest / bestsellers), price min/max, and pagination. Product takes an ASIN or full Amazon URL and returns title, brand, rating, buybox (price + availability), feature bullets, attributes, and variants. URL form auto-detects the storefront.

May 12, 2026 ​

  • Argus Owner Gets Ludicrous Mode, Group Members Get The Builder Bundle — owner Argus auto-merges every active tool in the catalog, no tool pre-selection needed; group/external senders now get a curated 33-tool allowlist covering research, analysis, and artifact generation (PDF/DOCX/XLSX/PPTX, code execution, image/music/video, finance and social search, medical_vision) while remaining locked out of owner data, owner state, and every Composio MCP.
  • Discord — Identity, Servers & Invites — connect your Discord account for a 10-tool read-only account surface: profile, server (guild) listing with member counts, invite resolution (codes, vanity codes, discord.gg URLs), widget/template inspection, third-party connection inventory, username/avatar management, and leave-guild. No channel messaging (Composio toolkit limitation); routes you to Slack/Teams/Gmail for that.
  • Argus Is A Full WhatsApp-Powered Agent — Argus now has a full operating guide, clearer setup path, autonomous heartbeat check-ins, model selection, expanded tool access, and use cases for running a capable agent from WhatsApp.

May 10, 2026 ​

  • Autonomous Workflows Are More Dependable — scheduled tasks run with fewer approval dead-ends, long-form briefs get expanded when too short, presentation creation retries drop, mobile chat controls stay in frame, and artifact outputs now end with concrete next-step offers

May 05, 2026 ​

  • Grok 4.3 Lands In The Picker — xAI's newest reasoning model is selectable now: 1M context, always-on reasoning, image input, tool calling, workhorse pricing; manual-pick only for now (not yet in Auto or Swarm)

May 04, 2026 ​

May 03, 2026 ​

April 30, 2026 ​

  • Qwen 3.6 Family Goes Live — Flash, 27B, Max Preview — three new Qwen 3.6 models in the picker today; Flash with 1M-token context (new tier), 27B dense reasoner, and Max Preview ~1T frontier MoE; three Qwen 3.5 entries retired with automatic autoupgrade for pinned sessions; every credit band stayed the same

April 29, 2026 ​

  • Argus On WhatsApp — dedicated Argus operating guide now live (activation flow, ACL/group mode rollout, sender trust boundaries, and whitelist/blacklist warnings), Tool Library now links directly to the Argus page, and the default server model for Argus sessions moves to Kimi K2.6

April 27, 2026 ​

April 25, 2026 ​

April 24, 2026 ​

April 23, 2026 ​

  • Medical Vision — AI Second Opinion On Your Scans (Local, Experimental) — new medical_vision tool runs MedGemma 4B locally via Ollama for X-rays, CT/MRI, ultrasound, dermatology, lab reports; inspect mode re-reads any image with forensic detail; 50k context window, pinned-warm residency, UI tagged Experimental, always disclaimed
  • Runway Gen-4.5 Takes Over Video Generation — video_lab now calls Runway Gen-4.5 for every clip (2-10s, preset aspect ratios, first+last keyframe, optional watermark); video_editor is fully NVENC-accelerated; retake/extend/upscale/ic_lora/id_lora/create_from_audio return temporarily_unavailable with migration guidance

April 21, 2026 ​

April 20, 2026 ​

April 19, 2026 ​

April 18, 2026 ​

April 17, 2026 ​

April 16, 2026 ​

  • Claude Opus 4.7 — Anthropic's frontier model with adaptive thinking, 1M context, swarm support

April 15, 2026 ​

Built for Alfrada OS.

Copyright 2026 Alfrada OS