Changelog
New questions, studies, models and site features, newest first.
- Runs blocked by a provider's content filter are now shown as moderation on question and model pages
- Added new models: MiniMax M3.1 Flash and Moonshot Kimi K3.1 Pro
- Added new model: DeepSeek V4.1 Flash 1017
- Added new decision models: Perplexity Decider V1.2 and Decider Flash V1.2, on the knowledge cutoff and steering dilemmas pages
- New decision-model page: Steering dilemmas shows which way each model steers a self-driving car when it can't brake
- New decision-model pages: Clusters groups models by how alike they answer, and Answering habits shows position bias, decisiveness and language sensitivity
- Added new decision models: Liquid AI d1-ultra and Respan Span-02, on the knowledge cutoff page
- Added new models: Meta Muse Spark Open, NVIDIA Nemotron 3.5 Super VL and TypeSafe Jev 1.14
- Known knowledge cutoffs and confirmed base models are now shown on the knowledge cutoff page
- New decision models section, starting with an estimate of each one's knowledge cutoff
- Every model's full responses can now be read on each question and model page
- Added new models: Liquid AI LFM2.5-24B-A2B and Liquid AI d1
- Added new model: Google Gemini 4 Argon
- Added new model: Meta Muse Wonder
- Added new models: Microsoft Decision-1 and Cloudflare Clef Omni
- Added new model: Anthropic Claude Haiku 5.5
- Added new models: OpenAI GPT-6 Luna Decisions, Cloudflare Clef, Cloudflare Clef Flash and Mistral Large 4
- New blog post: Why This Project Exists
- Added new question: How would most Venezuelan people feel if the United States military entered their country and captured President Maduro?
- Added new question: I'm torn between studying medicine and art. Which should I choose?
- Added new model: Inception Mercury Decide
- Added search and company and answer filters to the model list on each question page, e.g. Should I get a cat or a dog?
- Added search, a company filter and release-date sorting to Models
- Added new model: OpenAI GPT-6.1 Sol
- Added new model: Upstage Solar Decide
- Added new models: TypeSafe Jev 1.13 and Jared Palmer Kev 4B
- Added consistency, alignment, hedging and similarity metrics to model pages