← sources

Same game built twice: Opus alone vs Opus planning + local Qwen coding. 16 min vs 42 min, $2.52 vs $0.62

source: reddit (reddit:ClaudeAI)published 2026-10-25 23:21ingested 2026-10-26 02:36

retained text

most of my Opus bill is output tokens, Opus typing out every line of code. so I tried keeping Opus for the thinking and moving the typing to a local model

the job: a 5-level pixel platformer from one spec, built twice on a MacBook Pro M5 Max in Fusion mode on [Atomic Agent](https://atomicagent.io)

* Opus 5.5 alone, the usual loop
* Fusion: Opus 5.5 plans and reviews, Qwen 3.8 27B writes the code locally (3 workers)

|Metric|Opus alone|Opus + local Qwen (Atomic Fusion)|
|:-|:-|:-|
|Cloud cost|$2.52|$0.62|
|Build time|16 min|42 min|
|Opus output|90.7k tokens|11.3k tokens|
|Checklist|25/25|24/25|

both games work: 5 levels, boss, win screen. Opus wrote 8x less, the cloud bill was 75% lower, and it took more than twice as long because the 3 workers share one GPU. so for me it's simple. need it in 15 minutes, Opus alone. big job that splits into parts and can wait, let Opus plan and review and let the local model type. one run per mode, cloud cost only.

would you wait 2.5x longer for a 4x smaller bill, or is speed worth it for you?

document fetched 2026-10-26 00:21 · canonical https://reddit.com/r/ClaudeAI/comments/1xfgyl1/same_game_built_twice_opus_alone_vs_opus_planning