The verdict
Hammer AI's homepage claim — '100% free and completely private character roleplay' — is accurate, and the platform offers two distinct technical routes to local privacy. The more unusual one: WebGPU browser inference, where language models run directly in the browser tab without any installation, account, or local app — making Hammer AI one of the only tools that delivers client-side AI privacy from a standard web page. The second path uses Ollama or llama.cpp in the desktop app (Windows, Mac, Linux), for users who prefer a dedicated environment. Both routes mean prompts never reach Hammer AI's servers; 'your messages are not saved anywhere but your computer by default' is the exact stated promise. Twenty or more language models are supported across local and cloud paths. For users without suitable GPU hardware, a free daily cloud quota covers light use; paid cloud tiers (Starter ~$9/mo, Advanced ~$18/mo, Ultimate ~$35/mo) add higher context limits, image credits, and voice. Two creative modes run under one free plan: Characters (persistent companion roleplay with memory) and Stories (a standalone story-writing tool with Write Story, My Stories, and Leaderboard functions). Feature breadth on the free tier includes group chats with multiple characters simultaneously, AI image generation, TTS voice output, and Lorebooks/personas. The character filter covers 50+ specific tags — Female, Male, Non-binary, Human, Non-human, VTuber, Anime, RPG, Horror, Fantasy, and more. Platform reach: web, Windows, Mac, Linux desktop apps, plus iOS (App Store id6770699376) and Android (Google Play com.hammerai.app) — the iOS app was announced as just launched on the live homepage. The one honest constraint: local GPU performance — 7B models need 8 GB+ VRAM minimum.
Free, private, multi-platform local chat. For deeper model customization, SillyTavern; for polished cloud inference with GPU pricing, Backyard AI.
Pros
- WebGPU in-browser local inference — run AI locally from a browser tab with no install; privacy without leaving a web page
- Genuinely 100% free with no usage limits on local models (WebGPU browser + Ollama/llama.cpp desktop)
- Two modes: Characters (companion with memory) and Stories (co-authoring with dedicated Write Story tool)
Cons
- WebGPU browser path and Ollama desktop path both need a capable GPU (7B models need 8 GB+ VRAM minimum)
- Free cloud quota is metered — runs low in heavy sessions without a paid tier
- Less model variety and customization depth than SillyTavern for power users
Hammer AI FAQ
What is WebGPU inference on Hammer AI? It means Hammer AI can run language models directly in your browser using your GPU — no desktop app, no Ollama install, no account needed. Computation happens entirely on your machine, confirmed from the live site feature list as a distinct option from the Ollama desktop path.
What is the difference between Characters and Stories mode? Characters mode is persistent companion roleplay with memory — you build a relationship over time. Stories mode is collaborative narrative writing with a dedicated Write Story tool, separate from chatting with a companion.
Does Hammer AI have a mobile app? Yes. iOS (App Store id6770699376) and Android (Google Play com.hammerai.app) apps are available. The iOS app was announced as recently launched on the live homepage.
Is Hammer AI really completely free? Local use (WebGPU browser or Ollama/llama.cpp desktop) is fully free with no usage limit. Paid cloud tiers (Starter ~$9/mo, Advanced ~$18/mo, Ultimate ~$35/mo) add cloud LLM access, context window size, and image/voice credits for users without a GPU.
How many language models does Hammer AI support? 20+ language models confirmed from the live SoftwareApplication schema, spanning both local (on-device via WebGPU or Ollama) and cloud-hosted paths. A 'Local' filter tag in the character browser identifies characters optimized for on-device models.
Scored on the public six-point method, assessed against documented features, pricing and real user reports (plus hands-on use where we have account access) · last updated November 2026.