April 2026 · Powered by DeepSeek · Gemini · Groq · SambaNova · Cerebras

Your work,
amplified.

Memo AI is an intelligent work assistant built exclusively for Memo Fashion. Powered by DeepSeek V3.2 (685 billion parameters), Google Gemini 3, and 34 more AI engines — with live exchange rates, real-time weather, container tracking, image generation, and persistent memory that learns who you are across every conversation.

or

Only @memo.co.uk accounts are permitted

Online
Good morning, Memo
How can Memo AI help you today?
✉️ Write an email
🔍 Search the web
📊 Analyse data
💡 Brainstorm ideas
Ask Memo AI anything…
→
Production · April 2026

Enterprise AI. Multi-provider architecture.
Built for Memo Fashion.

Memo AI uses a cascade architecture — every message is routed through up to 14 AI engines across seven infrastructure partners. If one engine is at capacity, the next takes over in under 50 milliseconds. You never wait. You never see errors. The system adapts in real time.

0
AI engines
0
API connections
0
Max cascade depth
0ms
Fastest response
How the cascade works
When you send a message, Memo AI tries the best available engine first — like DeepSeek V3.2, a 685-billion parameter model. If that engine is busy, the system automatically switches to the next one (Groq GPT-OSS 120B, then Llama 4 Maverick, then Cerebras Qwen 3 235B...) — up to 14 engines deep. Each switch takes under 50 milliseconds. Multiple connections rotate across providers so no single endpoint is ever overloaded. The result: you always get an answer, always from a capable model, with zero downtime.
🌍
Live exchange rates
🌤
Real-time weather
📦
Container tracking
🧠
Persistent memory

Made for you.

Memo AI is built by the Memo internal team for the people of Memo Fashion. Here's what we made you.

✨
Pick your brain
Smart (DeepSeek V3.2 685B — beats GPT-4o on maths & code) · Reasoner (9-step deep thinking) · Live (Gemini + Google Search) · Fast (41ms first token) · Coder (DeepSeek V3.2) · Vision (SambaNova Llama 4 Maverick + Gemini). Each 4–11 engines deep.
→
⚙️
Under the hood
Full cascade architecture. 36-model matrix with latency benchmarks. FLUX.2 klein image gen & editing. DeepSeek V3.2 vs GPT-4o vs Claude comparison table. Daily limits per user. Data privacy & encryption details.
→
🛤
The journey
18 months of development. From a single AI model to 36 engines across 7 providers. Every release, every breakthrough — prototype to production.
→
🛠
Built in-house
Engineered by the Memo Internal Software Team. Not a ChatGPT wrapper. Not outsourced. Custom-built from zero lines of code to 36 AI models. 100% owned by Memo Fashion.
→
Six specialised models · multi-provider cascade

Pick the brain for the job.

Memo AI routes your question to the right model automatically. Each mode is backed by a cascade of 7–14 AI engines from Groq, SambaNova, Cerebras, Google and OpenRouter — so there's always capacity. Your files stay attached across messages. Generated images are stored permanently. And Memo AI remembers you — your name, your role, your preferences — across every conversation.

✨
Smart
1,000/day · 11-step cascade · 685B MoE
The everyday powerhouse. Polished emails, summaries, deep analysis, brainstorming, translation. Primary engine is SambaNova DeepSeek V3.2 — a 685-billion-parameter mixture-of-experts model that rivals GPT-4 on reasoning benchmarks. If busy, 10 more engines fire automatically.
Cascade (11 engines)SambaNova DeepSeek V3.2 (685B) · Groq GPT-OSS 120B · SambaNova Llama 4 Maverick · SambaNova DeepSeek V3.1 · Cerebras Qwen 3 235B · Groq Llama 3.3 70B · plus 5 OpenRouter :free models. Connections rotate automatically.
🧠
Reasoner
500/day · 9-step cascade · deep thinking
For problems that need real thought. Complex code, multi-step maths, legal reasoning, strategy. DeepSeek V3.2 leads — the same model that tops SWE-bench reasoning benchmarks. Shows its thinking process in a collapsible panel so you can follow the logic.
Cascade (9 engines)SambaNova DeepSeek V3.2 · DeepSeek V3.1 · DeepSeek V3.1-cb · Groq GPT-OSS 120B · Cerebras Qwen 3 235B · Llama 4 Maverick · Groq Llama 3.3 70B · plus 2 OpenRouter :free. Reasoning traces visible on demand.
🔴
Live
1,000/day · Gemini engines · real-time Google Search
Actually knows what happened today. Current news, weather, stock prices, sports scores, company info. Every answer is grounded in live Google Search results with sources cited at the bottom. Multiple Gemini connections across 3 models ensure consistent availability.
Cascade (4 engines)Gemini 2.5 Flash + Google Search · Gemini 2.5 Flash Lite + Search · Gemini 3 Flash Preview + Search · Groq GPT-OSS 120B + Tavily web search. Gemini engines + Tavily search = ~15,000 searches/day total capacity.
⚡
Fast
5,000/day · 2,000 tok/sec · 9-step cascade
Instant answers. Quick lookups, one-liner rewrites, “what's the word for…”. Cerebras Llama 3.1 8B runs at 2,000 tokens per second on WSE-3 hardware — the fastest public inference anywhere. Practically unlimited.
Cascade (9 engines)Cerebras Llama 3.1 8B (2,000 tok/sec) · Groq GPT-OSS 20B (41ms) · Groq Llama 3.1 8B · OR Nemotron Nano 9B :free · Liquid LFM 2.5 Thinking :free · GPT-OSS 20B :free · Gemma 3 4B · Gemma 3n 4B · Gemma 3n 2B.
💻
Coder
800/day · 9-step cascade · DeepSeek V3.2
Built for programming. Generates clean TypeScript, Python, SQL, HTML/CSS. Explains patterns, spots bugs, writes tests, refactors legacy code. DeepSeek V3.2 leads — the same model that topped the SWE-bench coding leaderboard.
Cascade (9 engines)SambaNova DeepSeek V3.2 · Groq GPT-OSS 120B · Cerebras Qwen 3 235B · DeepSeek V3.1 · Groq Qwen 3 32B · plus 4 OpenRouter :free (GLM-4.5 Air, GPT-OSS 120B, Qwen3 Coder, Arcee Trinity).
👁
Vision
500/day · auto-activates · image editing
Reads screenshots, photos, receipts, handwriting, diagrams, charts. Switches on automatically when you attach an image. Plus: paste any image and say “change to navy velvet” — FLUX.2 klein edits it with instruction-following AI, not random noise. Verified: colour changes actually work.
Cascade (8 engines)SambaNova Llama 4 Maverick · Gemini 2.5 Flash · Gemini 2.5 Flash Lite · Groq Llama-4-Scout 17B · Gemma 4 31B :free · Gemma 4 26B :free · Nemotron Nano VL :free · Gemma 3 27B :free. Image editing: Cloudflare FLUX.2 klein 9B/4B across 4 free accounts.
Capabilities · production-ready

Everything you need. Built in.

Every feature below is live right now. Sign in with your @memo.co.uk email and start using them immediately.

✍️
Professional Writing
emails · letters · reports
Write polished professional emails in seconds. Set the tone — formal, friendly, concise. Complete documents ready to send, not drafts with placeholders. Powered by DeepSeek V3.2 (685B params, outscores GPT-4o on writing benchmarks).
🔍
Live Web Search
news · prices · events · real-time
Ask about anything happening right now. UK minimum wage, weather in Milan, GBP to EUR rate, latest fashion week news. Grounded in live Google Search via multiple rotating Gemini connections. Sources cited at the bottom of every answer.
💱
Live Exchange Rates
ECB data · 13 currencies · instant
“Convert 5000 GBP to EUR” → instant answer with live European Central Bank rates. Supports GBP, USD, EUR, INR, CNY, JPY, AED, HKD, AUD, CAD, CHF, SGD, TRY. Powered by the European Central Bank.
🌤
Weather Anywhere
global · 3-day forecast · auto-geocode
“Weather in London” or “do I need an umbrella in Milan?” — current conditions, temperature, humidity, wind, UV index, plus a 3-day forecast. Any city in the world. Powered by Open-Meteo.
📦
Container Tracking
Maersk · MSC · CMA CGM · 6 more
“Where is my container MSCU1234567?” — generates instant tracking links for 9 major shipping lines: Maersk, MSC, CMA CGM, Hapag-Lloyd, COSCO, Evergreen, ONE, Yang Ming, ZIM. Built for fashion logistics.
🎨
Image Gen & Editing
FLUX.2 klein 9B · ~1.5s per image
Generate images from text: runway shots, tech flats, moodboards, product photography. Edit existing images with natural language — “change to emerald green” actually changes the colour (verified by a second AI model). Powered by Black Forest Labs FLUX.2 klein 9B, hosted on Cloudflare Workers AI.
📎
Document Analysis
PDF · Excel · Word · 10 files
Upload PDFs, spreadsheets, Word docs — up to 10 at once. Memo AI reads them, extracts the text, and answers questions about the content. Invoices, contracts, specs. Files stay attached across messages. Stored securely in Cloudflare R2.
🧠
Memory Across Chats
remembers you · 50 chats · auto-extract
Memo AI learns about you — your name, your department, your manager, your writing preferences. Facts are extracted automatically after each conversation and injected into every future chat. 50 conversations saved per person. The same memory capability as ChatGPT and Claude — built in.
Infrastructure · multi-provider architecture

The engine room.

Seven AI infrastructure partners. Multiple API connections. If one provider is at capacity, the cascade switches to the next in under 50 milliseconds. Staff never see an error — just a different engine label.

🟠
Groq
GPT-OSS 120B · Llama 3.3 · Qwen 3
Primary chat provider. Multiple connections rotating. GPT-OSS 120B (45ms), GPT-OSS 20B (41ms), Llama 3.3 70B, Qwen 3 32B, Llama-4 Scout (vision). Custom LPU chips — fastest inference hardware commercially available.
🟣
SambaNova
DeepSeek V3.2 · V3.1 · Llama 4
Frontier reasoning. DeepSeek V3.2 (685B MoE) is the primary Smart, Reasoner and Coder engine — outscores GPT-4o on MATH-500 (90.2% vs 76.6%) and HumanEval (92.7% vs 90.2%). Free tier, generous daily limits.
🔵
Cerebras
Qwen 3 235B · Llama 8B
World's fastest AI chip (WSE-3 wafer-scale). 2,000 tokens/sec on Llama 3.1 8B. Qwen 3 235B for heavy reasoning. Millions of tokens per day capacity.
🔷
Google Gemini
2.5 Flash · 3 Flash · grounded search
Live mode backbone. Multiple connections across Google Cloud projects. Gemini 2.5 Flash + Flash Lite + Gemini 3 Flash Preview. Google Search grounding built in — every Live answer is backed by real-time web results with cited sources.
🟢
OpenRouter
17 models · deep cascade fallback
Last-resort cascade. 17 verified working :free models including GPT-OSS 120B, Nemotron 120B/30B/9B, GLM-4.5 Air, Gemma 4 31B/26B, Gemma 3 27B/12B, Arcee Trinity, Liquid LFM. If every other provider fails, OpenRouter catches it.
🟡
Cloudflare
4 accounts · FLUX.2 klein · R2 storage
Image generation (FLUX.2 klein 9B & 4B), image editing (FLUX.1-schnell fallback), and persistent file storage in R2. 4 Cloudflare accounts provide ~400 images/day capacity.
🔴
Tavily
Structured web search · research-grade results
Fallback web search when Gemini is at capacity. Returns structured search results with titles, snippets, and URLs. Used by Live mode when Gemini fails, and by Smart mode for search-intent queries.