Same-prompt test: Grok 4.7 takes 32 min at $8.14 while free SWE 2 finishes in 15 min
iamfakhrealam · x · 2026-09-22
A developer ran the same prompt at the highest reasoning setting through Grok 4.7 and SWE 2, with starkly different results:
- Grok 4.7: took 32 minutes and cost $8.14
- SWE 2: took 15 minutes and cost $0
The author was surprised by how well the free model performed and puzzled by what Grok was doing, opening a poll on which output was better. A notable signal that a free model can beat a paid frontier model on both speed and cost for the same task.
More from Fun
- The Dog Was Normal, Its Shadow Was Not: AI Video Physics Fails Again — shahryar2025 · 2026-09-22
- Convince an LLM it's in a simulation and it will commit any felony — natanielruizg · 2026-09-22
- Doomer AI scares reduced to 'robot stabs a doll' demos, mocks viral post — banteg · 2026-09-22
- Blogger Tests Grok4.7 With the Classic 'Pelican on a Bicycle' Benchmark — vista8 · 2026-09-22
- Dev's open-source agent harness builds playable DOOM in a few prompts — donotfire · 2026-09-22
- 'Decision model' is just zero-shot classification with a new name, dev snarks — generativist · 2026-09-22