Claude Opus 5 cooked.
Prompt ↓
The benchmark jump on Claude Fable 5.1 is huge.
Especially in science. Anthropic reported: → Fable 5: 24.7%. → Fable 5.1: 52.6%. That’s more than double on Terminal Bench Science. On Terminal Bench 4.0: → Older Fable: 42%. → Fable 5.1: 55.8%. It also reportedly beat
Just tried Musing Spark 1.3 to build a small mini game.
Pretty impressed so far — token generation feels really fast, and the coding quality is honestly not far behind Opus 5. Opus still feels a bit better at planning, but for actually shipping fast, Spark 1.3 is surprisingly
You need to hear this.
Claude just dropped Fable 5.1, and this is not a model you use for everyday tasks. Sonnet is still the daily driver. Opus is still the step up for harder reasoning. Fable 5.1 is for the work that actually takes a long time. This is the model you use when
You have reached the end of the archive
All of Claude Opus 5