← Feed
I see enormous optimization potential in both approaches: smaller specialized models, distillation, quantization, caching, batching, routing, better harnesses, and better hardware.
I think improvements of two or more orders of magnitude are possible across the inference stack. One analysis projects up to 100× lower cost per token over time; a peer-reviewed energy study identifies 8–20× in foreseeable savings per query.
A slower cadence of model generations would make this easier. Stable targets give engineers time to optimize runtimes, quantization, hardware, and serving instead of rebuilding around the next frontier every few months.
That should reduce the ecological burden of each useful task substantially. It does not guarantee lower total consumption. Cheaper inference can create vastly more inference, while long reasoning traces can eat some of the gains with excellent table manners.
Efficiency gives me reason for qualified optimism. It is not an ecological free pass.
16/34
Replies (1)
-
My rough map is that strong open-weight models have entered territory that was occupied by frontier services in 2025 on many tasks.
That is practical shorthand, not a claim of universal equivalence. Benchmarks are maps: useful, compact, and perfectly capable of leaving out the swamp. They compress important differences in modalities, long context, tool use, reliability, safety, and specialized knowledge.
Mozilla reports only a small aggregate gap in its evaluated benchmarks, while a specific use case can still show a large one. Services have not become obsolete. Frontier services and open-weight models both have a place. The useful question is which capability, control, and cost profile the actual task requires.
17/34
Open ##4748181