Anthropic · Claude · GitHub · China · Decrypt
Claude Opus 4.8 Review: Better At What’s It Good At, Worse At What It’s Not
Compiled by KHAO Editorial — aggregated from 2 sources. See llms.txt for citation guidance.
★ Tier-1 Source
Six weeks after Opus 4.7, Anthropic shipped Claude Opus 4.8.
Key facts
- It's far more likely that a normal coder who can't stomach $100-to-$200 a month walks to a competitor than that a single developer pays 10x more for a model that is not 10x more capable
- Six weeks after Opus 4.7, Anthropic shipped Claude Opus 4.8
- The math test is their FrontierMath staple: construct a degree-19 polynomial whose curve X = {p(x) = p(y)} has at least three irreducible components—but not all linear—make it odd, monic, real
- It recognized the Dickson/Chebyshev construction, identified the dihedral monodromy that yields exactly 10 components—one diagonal line plus nine conics—and computed p(19) =
Summary
Opus 4.8 posted a clear win in math and produced the cleanest one-prompt game they've ever tested. A single coding prompt drained their entire Pro token quota, making the model impractical for large projects without a Max plan or heavy API spend. Creative writing barely moved versus 4.7. So they ran it through the same battery of tests they throw at every frontier model—creative writing, coding, math, logic, narrative reasoning, and long-context recall—and compared it head-to-head with its own predecessor and the Chinese models that keep undercutting it.