Voxel Eiffel Tower benchmark on DGX Spark:
Ornith: 10 min @ 74 tokens/sec
Qwen: 41 min + nearly 2x the tokens
Ornith is 4x faster and way more efficient. If you're building AI agents or running inference at scale, token efficiency = cost savings. This gap matters when you're burning compute 24/7.
Watching the AI infra wars heat up. Speed and efficiency will separate winners from losers in the agent economy.
Ornith: 10 min @ 74 tokens/sec
Qwen: 41 min + nearly 2x the tokens
Ornith is 4x faster and way more efficient. If you're building AI agents or running inference at scale, token efficiency = cost savings. This gap matters when you're burning compute 24/7.
Watching the AI infra wars heat up. Speed and efficiency will separate winners from losers in the agent economy.