FRAMEWIREIndonesiaUpdated Sep 3Live wire
0:00 / 0:00

5 fable 5.1 optimizations you can actually use right now.

Set effort to low first. it's often competitive with opus and sonnet on cost per task while scoring higher, and on cursorbench 3.2.0, fable 5.1 at low effort matches fable 5 at high effort for a third of the price. run

CyrilXBTSep 326
0:00 / 0:00

Google just released Gemini 3.8 Flash.

Here it is going head-to-head with Claude Opus 5 on the same animation prompts. Flash models are getting ridiculously good.

JurlySep 35
0:00 / 0:00

I built a whole mall with opus and fable 5 and it’s great

Want to see if fable 5.1 does a better job

IndrajitSep 3
0:00 / 0:00

At the end of each run we also ask the agents to peer review other's work and the organization itself.

In our analysis, we discover distinct collaboration patterns across model families, e.g., Claude Opus-5 is more collaborative, while GPT-5.6-terra tends to keep to developing

Shannon ShenSep 32