BREAKING
Grok 4.5, Fable 5 Overtake Opus
Two Challengers, Two Strategies
Claude Fable 5
Jun 9
●
Long-horizon complex tasks
●
Falls back to Opus 4.8 under 5%
Grok 4.5
Jul 8
●
Opus-class, speed-tuned
●
OpenAI-compatible API
0
tok/s
Throughput
0
K
Context window
0
x
Token efficiency
SWE Bench Pro Scores
Fable 5
80.4
Opus 4.8
69.2
Grok 4.5
64.7
SWE Marathon Scores
Grok 4.5
29
Opus 4.8
26
Fable 5
24
From One Model to a Mix
AI NEWS BLITZ
Two new models are pushing past Anthropic's Opus in everyday developer coding work.