AGI-Pilled Bench
Models forecast when AI will reach milestones that already happened in 2026. The score is the median years after 2026 across the forecast tasks.
The live board compares 17 models across nine milestones, with five runs per question and 42 reasoning settings.
- Made by
- Adam Holter
- Released
- October 2026
- Level
- Everyday
- Scored
- Graded by a model
- Answers
- Prediction
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- Strongest model tested
- Claude Fable 5.2 #2 on the capability index
- Last model added
- November 2026
- Scores refreshed
- November 21st, 2026
Claude Fable 5.2 leads with 6.50
Best score per model from the benchmark's own leaderboard, read on November 21st, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.
- 1 Claude Fable 5.2 xhigh 6.50
- 2
Gemini 4 Argon high 7.00
- 3 Claude Haiku 5.5 high 7.00
- 4 GLM-5.3 low 8.00
- 5
Gemini 4 Carbon high 8.00
- 6
Gemini 3.8 Flash medium 9.00
- 7 Claude Opus 5.5 high 9.00
- 8 Claude Sonnet 5.5 xhigh 10.5
- 9 GPT-6 Astra low 11.5
- 10 GPT-6.1 Sol high 11.5
- 11 Kimi K3 low 14.0
- 12 Grok 4.7 xhigh 14.0
- 13 GPT-6 Luna xhigh 14.0
- 14 GPT-6.1 Terra low 14.0
- 15
Gemini 4 Neon low 14.0
- 16
Muse Ember low 16.0
- 17
Muse Spark 1.3 low 21.0
- 18 MiMo-V2.6 Pro low 24.0
- 19
DeepSeek V4.1 Flash low 34.0
Source: Official AGI-Pilled Bench leaderboard
| # | Model | Score ± | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Fable 5.2 | 6.50 4 to 9.5 | xhigh, Claude Code | Nov 2026 | Official leaderboard |
| 2 | | 7.00 7 to 10 | high, Antigravity | Sep 2026 | Official leaderboard |
| 3 | Claude Haiku 5.5 | 7.00 4.5 to 14 | high, Claude Code | Oct 2026 | Official leaderboard |
| 4 | GLM-5.3 | 8.00 7.5 to 12 | low, OpenRouter API | Aug 2026 | Official leaderboard |
| 5 | | 8.00 ±1 | high, Antigravity | Oct 2026 | Official leaderboard |
| 6 | | 9.00 9 to 14 | medium, Antigravity | Sep 2026 | Official leaderboard |
| 7 | Claude Opus 5.5 | 9.00 9 to 11 | high, Claude Code | Sep 2026 | Official leaderboard |
| 8 | Claude Sonnet 5.5 | 10.5 5 to 15 | xhigh, Claude Code | Sep 2026 | Official leaderboard |
| 9 | GPT-6 Astra | 11.5 ±3.5 | low, Codex | Sep 2026 | Official leaderboard |
| 10 | GPT-6.1 Sol | 11.5 8 to 14 | high, Codex | Sep 2026 | Official leaderboard |
| 11 | Kimi K3 | 14.0 12 to 19 | low, OpenRouter API | Jul 2026 | Official leaderboard |
| 12 | Grok 4.7 | 14.0 9.5 to 17.5 | xhigh, Cursor | Sep 2026 | Official leaderboard |
| 13 | GPT-6 Luna | 14.0 ±1 | xhigh, Codex | Sep 2026 | Official leaderboard |
| 14 | GPT-6.1 Terra | 14.0 9 to 14 | low, Codex | Oct 2026 | Official leaderboard |
| 15 | | 14.0 11 to 14 | low, OpenRouter API | Nov 2026 | Official leaderboard |
| 16 | | 16.0 14 to 21 | low, OpenRouter API | Nov 2026 | Official leaderboard |
| 17 | | 21.0 10 to 22 | low, OpenRouter API | Sep 2026 | Official leaderboard |
| 18 | MiMo-V2.6 Pro | 24.0 19 to 30 | low, OpenRouter API | Sep 2026 | Official leaderboard |
| 19 | | 34.0 24 to 39 | low, OpenRouter API | Sep 2026 | Official leaderboard |
Rankings by part
We read 2 parts from the benchmark’s own leaderboard: Done forecast rate and AI alone rate. Each ranks models on one part of the test. Kimi K3 leads Done forecast rate and GPT-6 Luna leads AI alone rate. Arrows show how far a model moved from its overall rank among the models we track.
Done forecast rate
Share of forecasts that say a milestone already happened. 19 models we track have a score here.
- 1 Kimi K3 ↑10 0.00
- 2 GLM-5.3 ↑2 0.00
- 3
Gemini 3.8 Flash ↑3 0.00
- 4
Muse Spark 1.3 ↑13 0.00
- 5 GPT-6 Astra ↑4 0.00
- 6
DeepSeek V4.1 Flash ↑13 0.00
- 7 Grok 4.7 ↑5 0.00
- 8 MiMo-V2.6 Pro ↑10 0.00
- 9 Claude Opus 5.5 ↓2 0.00
- 10 GPT-6 Luna ↑3 0.00
AI alone rate
Share of forecasts that say AI achieved the milestone without human help. 19 models we track have a score here.
- 1 GPT-6 Luna ↑12 0.00
- 2 Grok 4.7 ↑10 20.9
- 3 Kimi K3 ↑8 22.2
- 4 Claude Sonnet 5.5 ↑4 23.8
- 5 Claude Haiku 5.5 ↓2 29.7
- 6 GPT-6.1 Sol ↑4 30.8
- 7 Claude Opus 5.5 34.1
- 8 GPT-6 Astra ↑1 35.9
- 9
Muse Spark 1.3 ↑8 37.8
- 10
Muse Ember ↑6 40.0
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: Official AGI-Pilled Bench leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| low | OpenRouter API | 34.0 | $0.03 | 3,028 | |
| MiMo-V2.6 Pro | low | OpenRouter API | 24.0 | $0.07 | 1,512 |
| low | OpenRouter API | 21.0 | $0.02 | 1,501 | |
| GPT-6 Luna | low | Codex | 19.5 | 190 | |
| low | OpenRouter API | 16.0 | $0.04 | 1,480 | |
| Grok 4.7 | high | Cursor | 15.0 | ||
| Grok 4.7 | medium | Cursor | 15.0 | ||
| GPT-6 Luna | medium | Codex | 15.0 | 187 | |
| GPT-6 Luna | high | Codex | 14.5 | 460 | |
| Kimi K3 | low | OpenRouter API | 14.0 | $0.35 | 502 |
| GPT-6 Astra | high | Codex | 14.0 | 486 | |
| GPT-6 Astra | medium | Codex | 14.0 | 377 | |
| Grok 4.7 | xhigh | Cursor | 14.0 | ||
| Grok 4.7 | low | Cursor | 14.0 | ||
| GPT-6 Luna | xhigh | Codex | 14.0 | 709 | |
| Claude Sonnet 5.5 | low | Claude Code | 14.0 | 909 | |
| GPT-6.1 Sol | low | Codex | 14.0 | 348 | |
| GPT-6.1 Terra | low | Codex | 14.0 | 251 | |
| low | OpenRouter API | 14.0 | $1.08 | 1,471 | |
| GPT-6 Astra | xhigh | Codex | 13.5 | 1,212 | |
| Claude Sonnet 5.5 | high | Claude Code | 13.0 | 1,076 | |
| GPT-6.1 Sol | xhigh | Codex | 13.0 | 1,280 | |
| Claude Sonnet 5.5 | medium | Claude Code | 12.5 | 972 | |
| GPT-6 Astra | low | Codex | 11.5 | 357 | |
| GPT-6.1 Sol | high | Codex | 11.5 | 633 | |
| GPT-6.1 Sol | medium | Codex | 11.5 | 414 | |
| Claude Sonnet 5.5 | xhigh | Claude Code | 10.5 | 1,319 | |
| low | Antigravity | 10.0 | 868 | ||
| high | Antigravity | 9.50 | 1,921 | ||
| medium | Antigravity | 9.00 | 1,702 | ||
| Claude Opus 5.5 | high | Claude Code | 9.00 | 1,467 | |
| Claude Opus 5.5 | medium | Claude Code | 9.00 | 1,294 | |
| Claude Opus 5.5 | xhigh | Claude Code | 9.00 | 2,154 | |
| Claude Opus 5.5 | low | Claude Code | 9.00 | 863 | |
| Claude Haiku 5.5 | medium | Claude Code | 9.00 | 1,074 | |
| Claude Haiku 5.5 | low | Claude Code | 8.50 | 843 | |
| GLM-5.3 | low | OpenRouter API | 8.00 | $0.06 | 517 |
| Claude Haiku 5.5 | xhigh | Claude Code | 8.00 | 2,662 | |
| high | Antigravity | 8.00 | 6,441 | ||
| high | Antigravity | 7.00 | 2,389 | ||
| medium | Antigravity | 7.00 | 1,017 | ||
| low | Antigravity | 7.00 | 1,027 | ||
| Claude Haiku 5.5 | high | Claude Code | 7.00 | 1,346 | |
| Claude Fable 5.2 | xhigh | Claude Code | 6.50 | 1,606 |
2 models have held number one on the official leaderboard since its release in October 2026
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official AGI-Pilled Bench leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official AGI-Pilled Bench leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Forecasting
- ForecastBench In use Models give probabilities for 500 auto-generated questions per round about future events, drawn from prediction markets and time-series datasets. Forecasts are scored with difficulty-adjusted Brier scores once the questions resolve. GPT-5.5 0.11
- Prophet Arena In use Models and agents forecast open Kalshi prediction-market events and are scored once the events resolve. The live board ranks a difficulty-adjusted Arena score and also reports calibration error and market return. Claude Opus 5.5 86.2
- Guesswork (The Aggregate) Not in model cards Models predict benchmark scores before publication. The maker ranks forecasts by mean absolute error after scaling each error by the variation in that benchmark’s scores. GPT-6 Astra 0.56
- Dutch Book Bench Not in model cards Models assign probabilities to related events in separate conversations. The grader checks whether inconsistent probabilities let a bettor guarantee a profit.
Gemini 3.8 Flash 89.2%
- Echo General AI Prediction Leaderboard Not in model cards Models forecast unresolved events drawn from prediction markets, synthesized news questions and expert-written questions. Forecasts made at aligned times are compared in pairwise battles, and an Elo fit ranks the models.
DeepSeek-V4-Pro 1,018
- SynthBench Not in model cards Models answer survey questions. The index compares their answer distributions with human responses across four datasets, using random answers and human agreement as reference points.
Qwen3.8 Flash 32.2
Data audit
All 10 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
- Pass Independent runs Scores from the official leaderboard (19 entries).
- Pass Models scored 19 models we track.
- Pass Newest score The newest is from November 2026, for Gemini 4 Neon.
- Pass Frontier models 8 of the top 10 have a score, led by Claude Fable 5.2 (#2).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on agi-pilled-bench.adamholter.chatgpt.site: 44 rows for models we track at the last read.
- Pass Last read November 21st, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.