Research Artifact
The Venus Elevation Map Claude Fable 5.1 Built
Most frontier-model claims end at a benchmark score. This one ends at a file: a new elevation map of a third of Venus, released under a Creative Commons license ahead of the next missions to the planet.
Key Facts
Page summary for citation
- Research Artifact: Most frontier-model claims end at a benchmark score. This one ends at a file: a new elevation map of a third of Venus, released under a Creative Commons license ahead of the next missions to the planet.
- Evidence trail: 2 sources, including Anthropic, Anthropic Docs.
- Internal context: related pages include GPU Kernels, Gallery, Verify.
"Claude Fable 5.1 trained a neural network to create a new, high-resolution elevation map of a third of the planet Venus."
What was built
Anthropic's Fable 5.1 announcement describes a digital elevation model of Venus produced by Claude Fable 5.1. The model trained a neural network on radar imagery from NASA's Magellan mission and on an existing elevation map that covered about one-fifth of the planet, then extended coverage to roughly a third of it.
The reported improvement is in resolution and accuracy rather than area alone: detail down to two to three kilometers instead of ten to twenty, and heights described as up to 25% more accurate. Anthropic says it released the map under a Creative Commons license ahead of the NASA VERITAS and ESA EnVision missions so that mission teams can use it when choosing observation targets.
Why it belongs in a build gallery
This site normally indexes software. The Venus map qualifies for the same reason a public repository does: there is a named artifact, a named producing surface, a license, and a stated intended user. A benchmark score cannot be inspected by a third party. A Creative Commons dataset can.
It is also a useful contrast with the Spotify, Ramp, and Rakuten case studies. Those are organizations reporting what Claude did inside their workflow. This one is the model vendor reporting what its own model produced, which is a weaker evidence position and needs to be labeled as such.
- Artifact: an elevation map released under a Creative Commons license.
- Surface: Claude Fable 5.1, generally available since 2026-09-01.
- Pattern: apply a model to public instrument data that nobody has had the budget to reprocess.
Evidence and limits
Every figure on this page comes from a single source: the Anthropic launch announcement for Claude Fable 5.1 and Claude Mythos 5.1. The resolution and accuracy claims have not been independently reproduced here, and this page does not evaluate the planetary science.
The honest framing is that this is a vendor-reported research artifact with a public deliverable attached. The deliverable is what raises it above a press claim, because the license means the map can eventually be checked by people who work on Venus for a living.
- Evidence tier: vendor announcement, not an independent benchmark or a peer-reviewed result.
- Not verified here: map accuracy, training method, or comparison against other elevation products.
- Watch for: use of the map by the VERITAS or EnVision teams, which would be much stronger evidence.
Primary Sources
Source trail
Links are cited for factual claims on this page. Access dates are kept so future updates can re-check drift.
Claude Fable 5.1 and Claude Mythos 5.1
Anthropic - official blog - accessed 2026-09-19
Fable 5.1 launch on 2026-09-01, default effort per surface, and the Venus elevation map, protein binder, and GPU kernel research artifacts.
Choosing a Claude model
Anthropic Docs - official docs - accessed 2026-09-19
Current generally available lineup and official guidance on picking between Fable 5.1, Opus 5, Sonnet 5, and Haiku 4.5.