Atlas

Benchmarks

Fiction.LiveBench

Fiction.LiveBench, published by Fiction.live, tests deep comprehension of long fiction: models answer private, open-ended quiz questions about character motivations, changing relationships, secrets and foreshadowing in stories cut to lengths from a relevant-passages-only version (0 tokens) up to 192K tokens. This suite-level row defines the family and has no direct score observations; Atlas records each context length on its own child benchmark.

Models
0
Top score
—
Leader
—
Updated
—
Resolution
—
Saturation
—

No signal or difficulty estimate. Held out of the fit by design.

This is a benchmark-family page. Scores are recorded on the component benchmarks below rather than directly on the parent, because the components use different score scales and directions.

MetricModelsLeaderDirectionDate
Fiction.LiveBench - 0 tokens (relevant passages only)54Opus 4100.0higher is better2025-09-29
Fiction.LiveBench - 16K tokens54Opus 4.6100.0higher is better2025-09-29
Fiction.LiveBench - 1K tokens54Opus 4.5100.0higher is better2025-09-29
Fiction.LiveBench - 2K tokens54Gemini 3 Flash Preview100.0higher is better2025-09-29
Fiction.LiveBench - 400 tokens54Opus 4.5100.0higher is better2025-09-29
Fiction.LiveBench - 4K tokens54Sonnet 4100.0higher is better2025-09-29
Fiction.LiveBench - 8K tokens54Opus 4.6100.0higher is better2025-09-29
Fiction.LiveBench - 32K tokens50Gemini 3 Flash Preview100.0higher is better2025-09-29
Fiction.LiveBench - 60K tokens50Gemini 3 Pro Preview100.0higher is better2025-09-29
Fiction.LiveBench - 120K tokens43Gemini 3 Flash Preview100.0higher is better2025-09-29
Fiction.LiveBench - 192K tokens25Gemini 3 Flash Preview100.0higher is better2025-09-29
Loading Atlas data…