From The Forge

News, releases, and notes from the BeastLab AI team.

· Open Source

Wisdom of Crowds for Your Codebase: Introducing the multi-llm Plugin

multi-llm is an open-source Claude Code plugin that fans a plan review, code review, or free-text question out to a panel of CLI-based LLMs in parallel, then consolidates their feedback for you to curate — the wisdom of a crowd of models without the copy-paste tax.

Read post →

· Announcements

Beast Models v3: URL Citations, Sharper Reasoning, and Improved Stability

Beast Nano, Beast Mini, and Beast Max move to v3 with URL citations returned in responses and shown in the Chat UI, improved reasoning in Beast Max, and better stability across the board.

Read post →

· Announcements

Beast Models v2: Prompt Caching, Sharper Vision, and More Reliable Tools

Beast Nano, Beast Mini, and Beast Max move to v2 with end-to-end prompt caching, more reliable tool calling, better image understanding, and cleaner API responses. See what is new, including the new cached-input pricing.

Read post →

· Showcase

Beast Max in One Shot: 3D Games From a Single Prompt

Seven self-contained Three.js builds — off-road racing, mech combat, demolition physics, and FPS scenes — each generated by Beast Max in a single shot. Play them, and read the exact prompts.

Read post →

· Announcements

BeastLab Unveils 3 New SOTA LLMs

Introducing three new multi-agent reasoning models — Beast Nano, Beast Mini, and Beast Max — spanning the full cost-to-capability spectrum, with Beast Max as our state-of-the-art flagship.

Read post →