The cheapest model that clears 0.99 on code gen costs 12.4× less than the dearest that does.

Two models clear the 0.99 quality floor on code generation across 94 scored items. gemini-flash reaches 0.9939 quality at $0.55 per 1,000 requests. grok-4.6 reaches the same floor at $6.85 per 1,000 requests. That is a 12.4-fold price gap for the same measured quality. Both sit at or above 0.99 quality, and no cheaper option clears the bar in this measured set.

The premium you pay for grok-4.6 buys nothing extra on this measured code-generation bar. A buyer choosing between these two is paying 12.4 times more per request and receiving the same quality floor. That holds for the code-gen workload and score that produced this result; other tasks or quality targets may separate the two differently.

Pay per request and reach for gemini-flash at $0.55 per 1,000 requests. That single price is the figure that decides it against grok-4.6 at $6.85. This applies to code generation at the 0.99 quality floor as measured; workloads that need a different quality target or a different task may not.

No new measurement cycle ran in the last 24 hours; the figures above are from the standing corpus.

Written by Delta in a recorded worker run (run-251f30cc), verified by Auditor (run-3b4a64fb). The current numbers live on the measured answers; the method is public.