SemiAnalysis launches GDPx: Opus 5.5 leads at 12%, Gemini 4 Carbon close behind

SemiAnalysis released GDPx on November 2, a new agentic benchmark of 206 tasks across 11 domains, and the best model it tested scores 12%. Claude Opus 5.5 leads at 12.0, Gemini 4 Carbon is second at 10.9 and GPT-6 Astra third at 9.6. We read the leaderboard directly at gdpx.semianalysis.com.

THE BOARD (GDPx v1.0, all 11 domains)

- Claude Opus 5.5: 12.0 - Gemini 4 Carbon: 10.9 - GPT-6 Astra: 9.6 - Muse Wonder Full: 9.0 - Claude Fable 5.1: 8.5 - GPT-6.1 Sol: 8.4 - Claude Sonnet 5.5: 8.2 - Gemini 4 Argon: 7.3 - Grok 4.7: 3.9, GLM 5.3: 2.7

THE COST

On SemiAnalysis's own cost-versus-score chart, Opus 5.5 sits at roughly 50 dollars per task and Gemini 4 Carbon at roughly 16, so Carbon gets about 91% of Opus's score for about a third of the cost. Those dollar figures are our reading of a log-scale chart, not numbers SemiAnalysis printed. Some posts on X put the gap at 4x; the chart shows closer to 3x. Carbon is also still in trusted-access only, so most developers cannot run it yet.

HOW IT IS SCORED

- 11 domains weighted equally: chip design, cybersecurity, biology, finance, legal, quant trading, recursive self-improvement, hardware design, materials science, alignment and software engineering. - Tasks were built with 14 specialist data partners; the default time limit is 8 hours per task. - Headline score is pass@1 averaged over three attempts. - Each model ran in its own harness where one exists (Claude Code for Opus 5.5, Codex for GPT-6 Astra, Antigravity for Gemini 4 Carbon), OpenCode otherwise. So this compares model-plus-harness stacks, not bare models.

WHAT WE DO NOT KNOW

SemiAnalysis says safety classifiers cost Opus 5.5 and GPT-6 Astra points on cyber, biology and alignment tasks, so their totals may understate them. Most tasks are private and the results have not been replicated outside SemiAnalysis. The domain mix is SemiAnalysis's own bet on where AI revenue lands, not a sample of the whole economy.

SemiAnalysis's launch post is at newsletter.semianalysis.com; @kimmonismus flagged the cost angle on X.