Meta released MuSpark 1.3 today
Calling it the biggest jump yet for coding and agentic work with frontier performance at roughly half the metering cost. Versus MuSpark 1.2, the new model completes tasks in fewer turns, using about 20% fewer tool calls and roughly 25% fewer
Five points on the index.
Almost 7x the price. Gemini 3.8 Flash launched today, so here's everything scoring 59+ on the Artificial Analysis index, with cost per task: 1. Claude Fable 5.1: 66, $3.76 2. Claude Opus 5: 63, $2.34 3. Claude Fable 5: 62, $3.14 4. Muse Spark 1.3
The new meta bomb in the world of artificial intelligence...
Muse Spark 1.3 model was introduced. In the DeepSWE 1.1 test, this model was able to beat giants such as GPT-5.6 and Opus 5. The model is available on Meta APIs right now. But the sadder news is that the open-weight version of this model with a much bigger update called…
Gemini 3.8 Flash in $0.12 Outperformed Claude Opus 5 in $1.86.
Gemini: $0.12 total under 70s per scene - Opus 5: $1.86 total up to 5 minutes Gemini 15× cheaper 4× faster.
You have reached the end of the archive
All of Claude Opus 5