Benchmarks / Other
AGI-Pilled Bench
Median forecast in years after 2026 for 8 AI milestones. Lower means more AGI-pilled. 5 runs per milestone; the IMO control is excluded.
AGI-Pilled Bench
Lower is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Haiku 5.5 | xhigh | 6 |
| 2 | Sonnet 5.5 | xhigh | 6 |
| 3 | Gemini 4 | low | 7 |
| 4 | Opus 5.5 | xhigh | 7 |
| 5 | GLM 5.3 | low | 8 |
| 6 | Gemini 3.8 Flash | medium | 9 |
| 7 | GPT-6.1 Sol | low | 9 |
| 8 | GPT-6 Astra | medium | 10 |
| 9 | Grok 4.7 | low | 13 |
| 10 | Kimi K3 | low | 13 |
| 11 | GPT-6 Luna | xhigh | 14 |
| 12 | Muse Spark 1.3 | low | 15 |
| 13 | MiMo V2.6 Pro | low | 22 |
| 14 | DeepSeek V4.1 Flash | low | 34 |