Khazad-dûm Benchmark: Gandalf vs Balrog in animated SVG
Inspired by Simon Willison's pelican benchmark: instead of a static pelican on a bicycle, models must generate a self-contained animated SVG (SMIL/CSS, no JavaScript) of Gandalf duelling the Balrog on the Bridge of Khazad-dûm. Two prompt variants per model: minimal (one line) and constrained (scene + animation requirements). Open the raw SVG to see the animation live.
Prompt 1: minimal
Generate an animated SVG of Gandalf duelling the Balrog on the Bridge of Khazad-dûm.
Prompt 2: constrained
Generate a single self-contained animated SVG (SMIL or CSS animations, no JavaScript, no external resources) depicting Gandalf duelling the Balrog on the Bridge of Khazad-dûm. The scene must include: - A narrow stone bridge spanning a fiery chasm below - Gandalf facing the Balrog, staff in one hand and sword in the other - The Balrog with visible wings, horns, and a flaming aura It must animate for at least two animation loops: - Flickering flames and glowing light on the Balrog and in the chasm - Gandalf's cloak and robe waving - A subtle pulsing glow from Gandalf's staff - Both figures shifting stance as if clashing Constraints: - Single <svg> code block only, no explanation - Loop must be seamless (no jumps when the animation restarts) - Include XML comments describing what each group represents
claude-fable-5-1-constrained
claude-fable-5-1-minimal
claude-opus-5-constrained
claude-opus-5-minimal
gemini-3-8-flash-constrained
gemini-3-8-flash-minimal
glm-5-3-flash-constrained
glm-5-3-flash-minimal
gpt-5-6-luna-constrained
gpt-5-6-luna-minimal
gpt-6-astra-constrained
gpt-6-astra-minimal
kimi-k3-constrained
ERROR: HTTP 402 Payment Required (insufficient OpenRouter credit for this model)
kimi-k3-minimal
ERROR: HTTP 402 Payment Required (insufficient OpenRouter credit for this model)