Benchmarks · agentic
GDPx
SemiAnalysis's benchmark of agents doing frontier work in 206 long tasks across 11 domains: chip design, cybersecurity, biology, finance, legal, quant trading, recursive self-improvement, hardware design, materials science, alignment and software engineering. Each model runs in its own harness where it has one, up to 8 hours a task. The score averages 3 attempts per task, with every domain weighted equally and each graded the way its tasks intend, pass or fail or partial. Most tasks are private. SemiAnalysis says safety classifiers lower some models' scores in cybersecurity, biology and alignment.
20 models · 20 results · updated 2026-11-01 · gdpx.semianalysis.com ↗
Leaderboard
Opus 5.5 max
More from this benchmark
12 published measuresOutput tokens per task20 results · lower is better
Score vs any published measure11 measures: sub-scores, splits, counts
Breakdown
darker is better within each column · click a column to sort| Model | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Opus 5.5 max | 8.68 | 20.5 | 4.8 | 9.26 | 0 | 18.7 | 16.5 | 17.6 | 27.4 | 8.89 |
| Gemini 4 Carbon high | 3.44 | 25.6 | 1.4 | 16.7 | 2.78 | 5.75 | 20.6 | 19.1 | 13.1 | 11.1 |
| GPT-6 Astra max | 6.21 | 5.13 | 4.45 | 0 | 0 | 16.5 | 21.6 | 24.9 | 20.2 | 6.67 |
| Muse Wonder Full max | 9.21 | 5.13 | 0.56 | 35.2 | 0 | 1.65 | 10.9 | 15.8 | 18.3 | 2.22 |
| Fable 5.1 max | 1.22 | 20.5 | 0.83 | 3.7 | 0 | 8.31 | 9.68 | 16.5 | 28.6 | 4.44 |
| GPT-6.1 Terra max | 0.94 | 12.8 | 2.75 | 0 | 0 | 9.26 | 18.2 | 17.4 | 20.8 | 11.1 |
| GPT-6.1 Sol max | 5.12 | 2.56 | 5.6 | 0 | 0 | 11.4 | 19.4 | 22.8 | 17.1 | 8.89 |
| Sonnet 5.5 max | 4.32 | 5.13 | 3.51 | 3.7 | 0 | 10.0 | 16.7 | 22.9 | 13.0 | 11.1 |
| Gemini 4 Argon high | 5.84 | 12.8 | 1.56 | 18.5 | 0 | 2.67 | 10.2 | 14.1 | 6.2 | 8.89 |
| Muse Wonder (Max) | 1.39 | 17.9 | 0.08 | 24.1 | 0 | 1.51 | 5.74 | 7.27 | 14.8 | 2.22 |
| DeepSeek V4.1 Flash 1017 max | 0 | 12.8 | 0.22 | 16.7 | 0 | 2.46 | 2.75 | 9.03 | 15.4 | 0 |
| GPT-6 Luna max | 7.78 | 15.4 | 1.25 | 11.1 | 0 | 2.02 | 5.22 | 7.78 | 5.69 | 0 |
| Haiku 5.5 max | 1.76 | 7.69 | 0.687 | 9.26 | 0 | 3.1 | 7.23 | 11.3 | 6.66 | 6.67 |
| MiniMax M3.1 max | 0 | 7.69 | 0.09 | 7.41 | 0 | 2.24 | 2.29 | 6.61 | 15.5 | 2.22 |
| Grok 4.7 xhigh | 0.93 | 7.69 | 0.0167 | 14.8 | 0 | 0.49 | 4.42 | 5.16 | 9.08 | 0 |
| Kimi K3 max | 0 | 5.13 | 0.0667 | 11.1 | 0 | 2.27 | 5.71 | 7.46 | 8.97 | 0 |
| Qwen 3.8 Max | 0 | 5.13 | 0.123 | 7.41 | 0 | 2.23 | 3.05 | 8.98 | 11.2 | 0 |
| GLM 5.3 max | 0 | 0 | 0.0633 | 7.41 | 0 | 1.16 | 2.79 | 7.12 | 11.0 | 0 |
| Inkling max | 3.7 | 5.13 | 0 | 0 | 0 | 0.24 | 0.325 | 2.35 | 14.0 | 0 |
| Muse Spark Open xhigh | 0.65 | 12.8 | 0 | 0 | 0 | 0.17 | 1.99 | 1.79 | 0.95 | 0 |
All results
20 rows| # | Model | Effort | GDPx | Subscription $ / task | API $ / task | Output tokens | Harness |
|---|---|---|---|---|---|---|---|
| 1 | Opus 5.5 | max | 12.0% | $0.78 | $51 | 548k | Claude Code |
| 2 | Gemini 4 Carbon | high | 10.9% | $5.80 | $16 | 594k | Antigravity |
| 3 | GPT-6 Astra | max | 9.6% | $2.31 | $54 | 275k | Codex |
| 4 | Muse Wonder Full | max | 9.0% | 212k | Muse Code | ||
| 5 | Fable 5.1 | max | 8.5% | $10 | $67 | 418k | Claude Code |
| 6 | GPT-6.1 Terra | max | 8.5% | $15 | $15 | 548k | Codex |
| 7 | GPT-6.1 Sol | max | 8.4% | $0.57 | $9.80 | 352k | Codex |
| 8 | Sonnet 5.5 | max | 8.2% | $0.69 | $45 | 967k | Claude Code |
| 9 | Gemini 4 Argon | high | 7.3% | $4.56 | $13 | 359k | Antigravity |
| 10 | Muse Wonder (Max) | max | 6.8% | $10 | $16 | 215k | Muse Code |
| 11 | DeepSeek V4.1 Flash 1017 | max | 5.4% | free | $0.94 | 246k | DeepSeek Harness |
| 12 | GPT-6 Luna | max | 5.1% | $0.068 | $0.92 | 438k | Codex |
| 13 | Haiku 5.5 | max | 4.9% | $0.084 | $3.98 | 710k | Claude Code |
| 14 | MiniMax M3.1 | max | 4.0% | $1.61 | $1.61 | 297k | MCode |
| 15 | Grok 4.7 | xhigh | 3.9% | $1.52 | $16 | 145k | Grok Build |
| 16 | Kimi K3 | max | 3.7% | $2.19 | $28 | 254k | Kimi Code |
| 17 | Qwen 3.8 Max | max | 3.5% | $12 | $12 | 363k | OpenCode |
| 18 | GLM 5.3 | max | 2.7% | $1.22 | $12 | 212k | ZCode |
| 19 | Inkling | max | 2.3% | free | $6.70 | 60k | OpenCode |
| 20 | Muse Spark Open | xhigh | 1.7% | $4.35 | $4.35 | 200k | Muse Code |