ChatLLM Web v3.1.0

v3.1 expands Local Model Studio into a device-aware gateway to the WebLLM model ecosystem.

Highlights

• 18 curated models across chat, coding, reasoning, vision, and local tool-use categories.
• 65 logical models backed by all 163 records in WebLLM 0.2.84.
• Advanced catalog switch for the complete official model directory.
• Logical model grouping keeps quantization and 1K variants under one catalog item.
• Automatic q4f16 or compatible q4f32 selection from reported WebGPU features.
• Stable, Experimental, Advanced, and Custom tiers.
• Recommended, Compatible, Cached, Coding, Reasoning, Vision, Tools, Experimental, and Custom filters.
• Search by model, family, capability, task, or official model ID.
• High-memory state based on declared VRAM and reported device memory.
• Active Advanced models remain available in the chat model picker.

Runtime

• WebLLM remains pinned to 0.2.84.
• One Web Worker and one active engine remain the runtime boundary.
• Every built-in runtime record is passed through the 4K application configuration.
• Existing conversations preserve their selected runtime model ID.
• The early handwritten inference layer remains archived under legacy/early-webllm-runtime.

Data and upgrade

v3.1 uses the existing chatllm-v3 IndexedDB schema. Conversations, attachments, preferences, custom models, and model caches remain available after the update.

Links

• Live: https://chatllm-web.pages.dev
• Repository: https://github.com/Ryan-yang125/ChatLLM-Web
• Models: https://chatllm-web.pages.dev/models