Epoch AI tests six models on real tasks; none meet its staff bar
11/17/2026 — 11/18, 06:18·1 sources (all)·1 reports (all)
Key takeaways
- Epoch AI ran six models, including Fable 5.2 and GPT-6.2 Astra, through its own real work tasks and found none met the bar it sets for emplo…
Background & analysis
On 2026-11-17, Epoch AI released the results of an internal evaluation in which it ran six models against the company's own real work tasks. The tested models included Fable 5.2 and GPT-6.2 Astra, along with four other unnamed systems. The assessment was designed around actual job tasks at Epoch AI rather than benchmark-style questions, and it was measured against what the organization describes as its employee standard, the level of performance expected of its own staff.
None of the six models met that employee standard, according to Epoch AI's findings. Notably, the two models that rank highest on Epoch AI's ECI index also fell short of the mark, which means top placement on that index did not translate into meeting the practical work benchmark used in the test.
As of the publication, Epoch AI has not disclosed the identities of the four remaining models it evaluated, nor has it released further details about the specific tasks or results. The available information therefore stops at the headline conclusion: across six systems, including the two ECI leaders, none reached the employee standard set by Epoch AI for real-world work performed at the organization.
AI-generated from 1 reports · updated 27 minutes ago
Related AI tools
- Heat index
- 52
- All sources
- 1
- All reports
- 1
- First seen
- 23 hours ago
Reports on this story headlines open the original
Epoch AI tested Fable 5.2, GPT-6.2 Astra and four other models on real work tasks drawn from its own operations. None met the bar its employees are held to, including the two models that rank highest on its ECI index. The evaluation used actual work rather than benchmark questions.
Epoch AIAI score 62
Other stories people are talking about
- Heat index 85NewOpenAI to deploy GPT-6 computer-use agents for US healthcare admin1 sources
- Heat index 363NewMicrosoft launches seven MAI models at Ignite 20264 sources
- Heat index 1006NewGPT-6.2 Astra and Sol Now Available on ChatGPT, API and Perplexity11 sources
- Heat index 96NewCalif says GPT-6 Cyber uncovered Java4Shell RCE flaw1 sources
- Heat index 91NewGPT-6 Cyber finds about 35 flaws in Apple products1 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion

