GDPx v1.0 - Materials Science - ValenceForge - Partial Score: leaderboard
GDPx v1.0 Materials Science - ValenceForge - Partial Score measures professional agent work using the publisher's native task-specific scores on a 0–100 scale. Three attempts are run per task, normally with an eight-hour limit. Scores mix binary and continuous grading; the overall mean weights eleven domains equally and partners equally within each domain. Native harness, effort and assigned/scored counts remain separate; missing native scores are not treated as zero.
Metric: Partial Score (%). Source: gdpx.semianalysis.com. Saturation forecast: Rough model projection: around 2026. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra (GDPx Run: v1.0; gpt-6-astra; Codex; effort Max) | 44.01 | |
| 2 | GPT-6.1 Sol (GDPx Run: v1.0; gpt-6-1-sol; Codex; effort Max) | 40.66 | |
| 3 | Claude Opus 5.5 (GDPx Run: v1.0; claude-opus-5-5; Claude Code; effort Max) | 38.68 | |
| 4 | GPT-6.1 Terra (GDPx Run: v1.0; gpt-6-1-terra; Codex; effort Max) | 37.4 | |
| 5 | Gemini 4 Carbon (GDPx Run: v1.0; gemini-4-carbon; Antigravity; effort High) | 35.66 | |
| 6 | Claude Sonnet 5.5 (GDPx Run: v1.0; claude-sonnet-5-5; Claude Code; effort Max) | 33.05 | |
| 7 | Gemini 4 Argon (GDPx Run: v1.0; gemini-4-argon; Antigravity; effort High) | 25.22 | |
| 8 | Claude Fable 5.1 (GDPx Run: v1.0; claude-fable-5-1; Claude Code; effort Max) | 23.71 | |
| 9 | Muse Wonder Full (GDPx Run: v1.0; muse-wonder-full; Muse Code; effort Highest) | 23.29 | |
| 10 | Claude Haiku 5.5 (GDPx Run: v1.0; claude-haiku-5-5; Claude Code; effort Max) | 11.62 |
Interactive version: theaggregate.ai/benchmark?slug=gdpx-v1-0-materials-science-valenceforge-partial-score · How It Works · Data refreshed daily, snapshot 2026-11-05.