Fetching from the wire…
01
02
03
04
05
06
07
08
Models2026-07-11 · source-backed
The July 10 refresh has Sol/Terra/Luna entering at 64.6/63.4/62.7% while Claude Fable 5 holds 80.3%, a +11.1 jump over Opus 4.8. Pro uses actively-maintained repos with no public ground-truth leakage, so its gap from the near-saturated Verified benchmark is the more honest signal. That ~16-point spread between top Claude and top GPT-5.6 on Pro is the number to watch when routing hard agentic tasks.
Each link below shares sources, entities, or timing with this story.
Copilot uses GPT / Shared entities / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Copilot uses GPT); both cover GPT, July, Luna, Opus; overlapping topics (between, claude, luna).
Gemini competes with Claude / Shared entities / Shared topic / Earlier coverage
Linked by a graph relationship (Gemini competes with Claude); both cover Claude, GPT, Opus, SWE; overlapping topics (benchmark, claude, gpt-5).
OpenAI released Terra / Shared entities / Shared topic / Earlier coverage
Linked by a graph relationship (OpenAI released Terra); both cover GPT, Opus, SWE, Verified; overlapping topics (benchmark, claude, gpt-5).