O SoloLLM 1.8 traz seus próprios modelos, conversas por voz e uma privacidade que você pode conferir. Os modelos continuam rodando no seu iPhone, e nada sai dele a menos que você escolha um recurso que acessa a internet. TRAGA SEU PRÓPRIO MODELO • Adicione modelos de chat GGUF das famílias Llama, Qwen, Gemma, Phi, Mistral e muitas outras: pesquise no Hugging Face dentro do app ou importe um arquivo do app Arquivos ou de um link. O SoloLLM verifica cada arquivo e mostra se ele cabe no seu aparelho antes de você adicioná-lo. • Os modelos importados rodam no seu aparelho como os incluídos, com um botão Report nas respostas. SEU PRÓPRIO SERVIDOR • Converse com modelos no seu computador: Ollama, LM Studio ou qualquer servidor compatível com OpenAI. Sem conta. O SoloLLM pergunta antes de enviar o primeiro chat a cada servidor. FALE COM SEU MODELO • Conversa por voz sem usar as mãos: toque no botão de onda ao lado da caixa de mensagem, fale e ouça a resposta. Fale por cima para interromper. O reconhecimento de fala acontece no seu aparelho. • Comece com "Talk to SoloLLM", o Action button (botão de Ação) ou um controle da Central de Controle. PRIVACIDADE QUE VOCÊ PODE CONFERIR • Offline only, em Settings > Privacy & Data, bloqueia todos os acessos à rede que o SoloLLM faz. • O Network Log mostra cada acesso que o SoloLLM fez: para onde foi, por quê e quando. • Uma etiqueta abaixo de cada resposta mostra onde ela foi escrita: no seu aparelho, no seu servidor ou em um provedor de nuvem. • Novas instalações começam com a sincronização de conversas com o iCloud desativada. Se você já usa o SoloLLM, sua configuração continua igual. O iCloud Backup está pronto em Settings > Cloud & Sync. CONVERSAS QUE CONTINUAM • Conversas longas não esbarram mais em um limite. As mensagens antigas são resumidas no seu aparelho, um medidor mostra o espaço que resta e você pode definir o tamanho do contexto. ATALHOS • Nova ação Run a Task, mais receitas: Summarize Voice Memo, Receipt to Table e Summarize Clipboard. No iOS 27, modelos menores podem executá-las sem abrir o app. MODELOS E FERRAMENTAS • Novos modelos: Granite 4.2 3B, MiniCPM-V 4.7 1B para fotos e Gemma 4 12B para iPads com 12 GB de memória ou mais. • Chamadas de ferramentas mais confiáveis: cada família de modelos usa seu próprio formato, e as chamadas são verificadas enquanto o modelo as escreve. • Motor llama.cpp atualizado. • Uma única verificação de memória decide quais modelos cabem no seu aparelho, e a configuração inicial sugere um para começar. • Alguns modelos antigos não são mais oferecidos para download. As cópias que você já tem continuam funcionando. • Servidores MCP e ferramentas HTTP personalizadas têm sua própria tela, Tools & Connections, em Settings > Agent & Tools. Ferramentas que usam seu calendário, lembretes, câmera, área de transferência ou arquivos começam desativadas. • Fixe conversas no topo e arquive as antigas. CORREÇÕES • O MiniCPM5 1B voltou a responder corretamente. • Corrigido um travamento nas ferramentas de reconhecimento de fotos e de texto.

A big reliability update. IMPROVED • Chat: Stop is cleaner, a failed reply is marked as failed, and a partial local reply is kept when you leave the app. • Gemma 4 replies no longer show its internal reasoning or tool-call markup. • History search finds text inside messages, not just titles. • Downloads resume after an interruption, and each model variant shows its own progress. • Knowledge indexes whole PDFs, and documents indexed by older versions are flagged for re-indexing. • Math renders offline and follows light and dark mode. • Dictation and read-aloud share audio smoothly, and read-aloud skips markdown symbols. • Cloud Mode asks before sending anything, retries brief network failures, shows the provider's error message and never hides a cut-off reply. • Agent tools: approve tools one at a time, turn off any built-in tool, and a safer web fetch and code interpreter. • Shortcuts and widget prompts reach a running app and wait for the model to load. • iPhone Duo: Models, Settings and Templates get a sidebar and a comfortable width when unfolded, and open sheets stay open across a fold. • Better layouts at the largest text sizes. FIXED • Photos and saved prompts stay with their conversation after Start Fresh, regenerate and resume. • A storage problem now shows a screen you can recover from. • API keys and your iCloud choice survive Reset to Defaults.

Ready for iPhone Duo. • SoloLLM now fills both Duo screens and adapts as you fold and unfold. • Folding mid-reply keeps your conversation, your draft and the reply in progress, plus any attachment, open sheet or download. • Wider screens get a comfortable reading width for messages and the composer. • More reliable storage on iOS 27.

Ask questions about your own documents, save your favourite prompts, and use SoloLLM from Siri and Shortcuts. Everything still runs on your device. NEW - Knowledge (beta): add PDFs, text or Markdown files in Settings > Knowledge, then ask the model about them. Files are indexed on your iPhone and never leave it. - Prompt library: save your own prompts in Templates, with {{blanks}} you fill in each time. - Siri and Shortcuts: Ask, Quick Answer, Summarize, Extract Text and Translate. - MiniCPM5 2B: a new 1.5 GB model that is good at tools and code. - Wi-Fi only downloads, on by default. - Cloud Mode: use your own Mac or PC server, pick its model by name, and long replies no longer time out. FASTER AND SMOOTHER - Long replies stream with far less work for your phone, and the chat screen no longer redraws on every word. - A cleaner first launch that takes you straight to the models. - Haptics on send, stop and copy, and a calmer, more consistent look across the app. FIXES - Fixed a crash when tapping the microphone. - Emoji, accents and Japanese, Chinese or Korean text no longer turn into boxes. - A model's private reasoning no longer spills into its answer. - Downloads explain what went wrong in plain words and check free space first. - The iCloud sync switch now really turns sync on or off. - Web pages can no longer smuggle instructions to the model.

The newest on-device models, a faster engine, Apple Intelligence, and image input. NEW MODELS Qwen3.5 (0.8B to 9B), LFM2.5 2.6B and 8B-A1B, Nanbeige4.2 3B, MiniCPM5 1B, and Google's official Gemma 4 QAT builds. Qwen3.5 4B is the new recommended default: 262K context, thinking mode, and tool calling. Every download is verified against a pinned digest. APPLE INTELLIGENCE On iOS 26 with Apple Intelligence enabled you can start chatting immediately with Apple's on-device model, no download required. Still fully private. IMAGE INPUT Vision-capable models (Qwen3.5, Gemma 4) can now look at your photos directly instead of reading extracted text. Download "vision support" from a model's detail page. FASTER ENGINE Updated llama.cpp with per-device Metal flash attention, a q8_0 KV cache option that fits twice the context in the same memory, and warm-up on load so the first reply starts sooner. Thinking off now works on every reasoning model. A REFRESHED LOOK A sleeker design with Liquid Glass on iOS 26, a cleaner chat screen, and clearer model cards.

Privacy & stability improvements. All 17 models now have verified integrity checksums — downloads are validated on-device before the model activates. Fixed a bug where network-tool consent state wasn't saved correctly across sessions. Widget now shows the accurate downloaded model count. Improved privacy disclosures throughout the app.

Stability improvements

Version 1.3 — The biggest update yet. NEW MODELS • Qwen 3 (0.6B, 1.7B, 4B) — Alibaba's latest generation with built-in thinking mode. Watch the AI reason through problems step by step before answering. Toggle thinking on or off per message. • Google Gemma 4 E2B — Ultra-efficient 2B model that packs surprising capability into under 1 GB. Our new recommended pick for most users. • Google Gemma 4 E4B — More horsepower when you need it. 4B parameters with native tool calling for agent mode. THINKING MODE Qwen 3 models can now show their reasoning. Tap the brain icon in the toolbar to enable thinking mode — the model works through the problem internally before giving you an answer. Great for math, logic, and complex questions. LIGHT & DARK MODE Settings → Appearance now offers Light, Dark, or follow the system. The default stays dark, but Light mode is finally here for those who prefer it. AGENT MODE (MAJOR UPDATE) The built-in AI agent has been completely overhauled: • 23 native tools — web search, web browsing, OCR, camera, image analysis, code interpreter, PDF reader, document reader, file reader/writer, clipboard, translation, URL dispatch, data analyzer, scratchpad, memory, plan mode, and more • Approval mode — choose whether tools run automatically or need your sign-off first • Verbose mode — see exactly what the agent is doing and why • Max steps cap — set how many tool calls the agent can make before wrapping up • MCP server support — connect to local MCP servers on your home network (Home Assistant, custom tool bridges, etc.) • Agent Live Activity — track agent progress right from your Lock Screen or Dynamic Island TEMPLATES Tap the template icon to jump-start common tasks: • Scan a Receipt — point your camera at a receipt, extract all line items, export to CSV or add totals to Reminders • Business card scanner, photo description, whiteboard capture, plant and food identification, fix my writing, email draft, and more MODEL BROWSER IMPROVEMENTS • Three clear states — Active (loaded and running), Downloaded (on device, not loaded), and Not downloaded • Unload button — free up RAM without closing the app • Swap confirmation — clear dialog when switching models so you never accidentally interrupt a session • Checkmarks now refresh immediately after deleting a model EDIT & EXPORT • Edit any sent message and regenerate the response from that point • Export the full conversation as JSON STABILITY Dozens of fixes across this release, including stop-race guard (no more duplicate responses), Vision OOM prevention for large images, auto-load RAM gate (skips auto-load if available memory is too low), widget sync fixes, keychain credential update, and much more.