What's new at League of LLMs.

• Added Meta Muse Spark 1.2 as a paid candidate model
• Added native Meta web search and source citations
• Added native image, PDF, audio, and video input for Muse Spark

Muse Spark 1.2 now uses Meta's Responses API with streaming, native files, automatic prompt caching, and the arena's existing search-decision harness.

What this means for you: Scout, Arena, and Champion members can include Muse Spark in the same comparison and judge evaluation as every other candidate response.

• Added Free, Scout, Arena, and Champion usage plans
• Added server-side usage reservations and file-upload quotas
• Added Stripe-hosted subscription checkout and billing management

Usage is now estimated before submission and enforced transactionally on the server, with plan-specific model, judge, usage, and file limits.

What this means for you: You can try ten prompts without an account, use a daily Free allowance, or choose a paid plan for all models and higher limits.

• Added Anthropic Claude Sonnet 5 as a candidate and judge provider
• Added native Claude image, PDF, and text-document inputs
• Connected Claude to the shared live-search grounding flow

Claude Sonnet 5 now participates in the same side-by-side arena, file workflow, search grounding, and judge pipeline as every other provider.

What this means for you: You can compare nine providers in one query and include Claude in both candidate and judge roles.

• Added Groq and GLM as supported providers
• AI judge now highlights the best response with an explanation
• PDF uploads now supported alongside images

This release significantly expands the range of models available in the arena and makes the AI judge far more useful. It now tells you not just which response won, but why.

What this means for you: You can now compare 7 models in a single query, upload PDFs directly, and get a clear explanation of which answer is best.

• Launched side-by-side comparison workspace
• Added support for Hugging Face and Cohere
• Improved response loading speed

The core arena workspace landed in this release: the side-by-side layout that makes comparing model outputs fast and visual rather than a copy-paste exercise.

What this means for you: All model responses now load in parallel in the same view, so you never have to switch tabs.

• Initial launch of League of LLMs arena
• Support for OpenAI, Gemini, and Mistral
• Image upload support (JPG, PNG)

The first public version of the arena set the starting point for everything: three models, image support, and the core comparison idea proven out.

What this means for you: You can compare OpenAI, Gemini, and Mistral on any prompt and attach images to your queries.