Kevin Bryan's HardGeoBench: Fable 5.2 ends Gemini's 18-month reign at photo geolocation
Afinetheorem · x · 2026-11-10
Kevin Bryan (U Toronto / All Day TA) launched HardGeoBench, a raw-LLM benchmark testing frontier models on pinpointing photo locations with no metadata, text, or GeoGuessr-style clues — never posted online, no web search allowed, pass@1 with a fixed prompt.
On his combined logic+image benchmark, Fable 5.2 tops the board for the first time in 18 months, dethroning Gemini. Scores at ≤10-mile precision: Fable 5.2 (49), Gemini 3.8 (46), Muse Spark Open (39), Astra 6.1 (38), Kimi k2.5 (30), GLM5.3 (28), Q3.8Max (23), Grok 4.6 (20), DS v4FV (15). Across three re-runs of the top models in May 2025, at least one nailed the exact location 57.5% of the time. Bryan notes the benchmark also doubles as a way to probe training-set contamination.
More from Models
- Fable 5.2 reportedly one-shots benchmarks 'nothing came close to' — EricBuess · 2026-11-10
- How AA-3D Works: Models as Agents in Sandboxes Building 3D Scenes — ArtificialAnlys · 2026-11-10
- Two Generations Compared: Claude Fable 5.2 vs 5.1 on the Same 3D Brief — ArtificialAnlys · 2026-11-10
- AA-3D Briefs Explained: From Starship Missions to Taj Mahal Walkthroughs — ArtificialAnlys · 2026-11-10
- Claude Fable 5.2 Tops AA-3D Benchmark With 1271 Elo, 67 Points Ahead — ArtificialAnlys · 2026-11-10
- Dev builds a playable browser game with Claude Fable 5.2: sail through 8 dream worlds — FinanceYF5 · 2026-11-10