Epoch Automation Reports
We show this table for reference; we do not rank on it.
Epoch Automation Reports runs AI agents on open-ended tasks taken from Epoch AI's own work, such as graphic design, data insights and AI data center research, and has human graders score the output against Epoch's internal standards.
Benchmark score on Epoch Automation Reports — November 13, 2026
We compile the Epoch Automation Reports rows from benchmark-owner or independent runs. Claude Fable 5.1 leads the table at 65.3%, followed by GPT-6 Astra (65.1%) and Grok 4.6 (59.1%). We do not use these results to rank models overall.
Claude Fable 5.1
Anthropic
Run by Epoch Automation ReportsGPT-6 Astra
OpenAI
Run by Epoch Automation ReportsGrok 4.6
xAI
Run by Epoch Automation Reports6 modelsAgenticCurrentDisplay onlyUpdated November 13, 2026
Benchmark score table (6 models)
| Rank | Model / configuration | Score | Who ran it | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Claude Fable 5.1AnthropicRun by Epoch Automation Reports | 65.3% | Run by Epoch Automation Reports | Not reported | Closed |
| 2 | GPT-6 AstraOpenAIRun by Epoch Automation Reports | 65.1% | Run by Epoch Automation Reports | Not reported | Closed |
| 3 | Grok 4.6xAIRun by Epoch Automation Reports | 59.1% | Run by Epoch Automation Reports | Not reported | Closed |
| 4 | Qwen3.8 MaxAlibabaRun by Epoch Automation Reports | 53.0% | Run by Epoch Automation Reports | Not reported | Open |
| 5 | Kimi K3Moonshot AIRun by Epoch Automation Reports | 52.4% | Run by Epoch Automation Reports | Not reported | Open |
| 6 | Gemini 3.8 FlashGoogleRun by Epoch Automation Reports | 42.3% | Run by Epoch Automation Reports | Not reported | Closed |
- Lab-reported
- Benchmark-run
- Independent
How to read this leaderboard
Higher is better. Claude Fable 5.1 in Claude Code (65.3%) and GPT-6 Astra in Codex (65.1%) are effectively tied at the top. A score is for a model plus its agent harness, not the model alone.
Operator receipt: 6 sourced rows are currently displayable on this page; the leading published row is Claude Fable 5.1 at 65.3%.
Honest limit: Six agent systems and 11 tasks, graded by humans, so individual scores carry judgment and small-sample noise. The tasks reflect one research organization's work. Epoch publishes this board only on its page, not in its data download. BenchLM shows it as display-only; it does not enter overall or category rankings.
Among the reported Epoch Automation Reports rows, Claude Fable 5.1 is first at 65.3%. The third row is 6.2 points behind. The broader top-10 range is 23.0 points, so the table still separates the published systems.
6 models have been evaluated on Epoch Automation Reports. The benchmark falls in the Agentic category. We keep these results separate from the weighted global scoring system, so they remain source-specific evidence. Epoch Automation Reports is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About Epoch Automation Reports
Year
2026
Tasks
11 open-ended work tasks in five categories from Epoch AI's own work
Format
Agent harness at highest reasoning setting, human rubric grading
Difficulty
Open-ended professional research and design work
The task suite currently has 11 tasks in five categories: graphic design, data insight generation, data explorer generation, AI data center research and research design. Each model runs in its own agent harness (for example Claude Code, Codex or Grok Build) at its highest reasoning setting, with the context a human contractor would get. Memory and skill writing are disabled so nothing carries over between tasks. A human grader scores each output against rubrics that reflect Epoch's quality bar; 100% means the work meets the standard Epoch expects from its own staff. BenchLM stores the average across categories and lists the category scores in each row's notes.
Freshness and provenance
Version
Epoch Automation Reports 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What are Epoch Automation Reports?
A benchmark of real tasks from Epoch AI's own work, from graphic design to AI data center research. Agents produce the work and human graders score it against Epoch's internal standards.
Which AI agent scores best?
Claude Fable 5.1 in Claude Code (65.3%) and GPT-6 Astra in Codex (65.1%) lead, followed by Grok 4.6 in Grok Build (59.1%). Gemini 3.8 Flash in Antigravity is last at 42.3%.
Compare top models on Epoch Automation Reports
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.