AI model task length scaling
METR's scatter plot drew 3.6M views on X: 23x its typical post and #3 of 26 METR charts we track by views. Read Oct 10, 2026.
We estimate that Claude Opus 4.6 has a 50%-time-horizon of around 14.5 hours (95% CI of 6 hrs to 98 hrs) on software tasks. While this is the highest point estimate we’ve reported, this measurement is extremely noisy because our current tas
Why it spread
This post spread because it paired a striking headline number, a 50%-time-horizon of around 14.5 hours with a 95% CI of 6 hrs to 98 hrs, with unusually candid framing that the measurement is extremely noisy and the task suite is nearly saturated, drawing 3,972 likes, 440 reposts, 477 quotes, 1,232 saves and 3,552,976 views.
What worked
- The headline number is concrete and easy to repeat, around 14.5 hours with a 95% CI of 6 hrs to 98 hrs, so people could quote it directly.
- The post admits the weakness of its own result, saying the measurement is extremely noisy and the task suite is nearly saturated, which builds trust and invites discussion.
- The chart shows task length at 50% success growing exponentially over model releases, a simple visual story that matches the text and is highly self-labelled and branded with a source line.
What to borrow
We could use the same form by leading with one clear number from our own data, showing the trend as a simple scatter, and stating the uncertainty and limitations in the post text instead of hiding them.
Careful
Do not copy the exact numbers or treat the point estimate as precise, since the post itself says the measurement is extremely noisy and the task suite is nearly saturated, and do not over-read the high quote rate as agreement.
Written by our analysis model from the measured numbers above, not by METR.
Make it yours
Your numbers, in this line chart.
Starts from the real numbers of "Task length (at 50% success time-horizon) vs Model release date" by METR. Replace every number with yours before you post. Make one like this, step by step
Free pictures carry a small "Made with ViralChart" line. That line is how other people find us.
API · Max
Use this chart via API
On the Max plan you can pull this chart's numbers with the provenance of each one, as JSON or CSV, with its redraw spec, or call it from an agent over MCP. Chart id bcb28382a056.
| Path | Returns | Plan | |
|---|---|---|---|
| GET | /api/v1/charts/bcb28382a056 | Title, publisher, date, form, point count; with a Max key, the data link | Open |
| GET | /api/v1/charts/bcb28382a056/data | Every point with its provenance and confidence, plus the redraw spec. Add ?format=csv for CSV | Max |
| GET | /api/v1/charts/bcb28382a056/render?view=1 | The draw-only view this page uses | Open |
| POST | /mcp | Tool get_chart_data with this id, for Claude, Cursor and other agents | Max |
curl -s https://viral-chart.com/api/v1/charts/bcb28382a056/data \
-H "Authorization: Bearer $VIRALCHART_KEY"claude mcp add --transport http viralchart https://viral-chart.com/mcp \
--header "Authorization: Bearer $VIRALCHART_KEY"60 requests a minute per key; over that you get 429 with Retry-After. Up to 5 keys per account, made on your account page and shown once. Without Max, the data route answers 402 with a link to the plans. The numbers describe a chart its publisher owns, and every file carries a license note.
More from METR
All 26
Other giants using a scatter plot
All 78
Questions people ask
How far did METR's "AI model task length scaling" chart travel?
3,552,976 views, 3,972 likes, 440 reposts and 1,232 saves on X, read Oct 10, 2026. A normal post from @METR_Evals gets about 153K views.
What kind of chart is it?
A scatter plot. Task length at 50% success grows exponentially over model releases.
Can I use METR's chart?
Use the format, not the picture. The chart, logo and design belong to METR. ViralChart redraws the chart type with your numbers and words, and credits the original publisher.