LisanBench
We show this table for reference; we do not rank on it.
A word-chain reasoning benchmark that tests planning, recall, constraint following, and vocabulary depth by asking models to extend non-repeating edit-distance-1 chains.
Difficulty-weighted score on LisanBench — September 23, 2026 snapshot
We mirror the published difficulty-weighted score view for LisanBench. Claude Opus 4.7 (Adaptive) leads the public snapshot at 5122.60, followed by Claude Opus 5 (5068.25) and Claude Fable 5 (4561.82). We do not use these results to rank models overall.
Claude Opus 4.7 (Adaptive)
Anthropic
Claude Opus 5
Anthropic
Claude Fable 5
Anthropic
154 modelsReasoningCurrentDisplay onlyUpdated September 23, 2026 snapshot
Difficulty-weighted score table (154 models)
ScoreHow LisanBench is shown here
BenchLM mirrors the current public LisanBench difficulty-weighted leaderboard using the official dataset published at lisanbench.com for the September 23, 2026 snapshot. The public benchmark tests 154 model variants across 50 starting words, with 3 trials per starting word.
LisanBench is a strong reasoning reference, but BenchLM currently keeps it display only rather than weighted. The public leaderboard is highly variant-specific, strongly English-vocabulary-dependent, and not yet aligned cleanly enough with BenchLM canonical model rows to use as a ranking input.
Snapshot
The published LisanBench snapshot places Claude Opus 4.7 (Adaptive) first at 5122.60. The third row is 560.78 score units behind. The broader top-10 range is 2852.20 score units, so the table still separates the published systems.
154 models have been evaluated on LisanBench. The benchmark falls in the Reasoning category. LisanBench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About LisanBench
Year
2026
Tasks
50 starting words × 3 trials
Format
Difficulty-weighted word-chain reasoning
Difficulty
Open-ended lexical planning
BenchLM mirrors the public difficulty-weighted LisanBench leaderboard as a display-only reasoning benchmark. The public benchmark currently evaluates 128 model variants across 50 starting words with 3 trials per word.
Freshness and provenance
Version
LisanBench 2026
Refresh cadence
Static
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does LisanBench measure?
A word-chain reasoning benchmark that tests planning, recall, constraint following, and vocabulary depth by asking models to extend non-repeating edit-distance-1 chains.
Which model leads the published LisanBench snapshot?
Claude Opus 4.7 (Adaptive) currently leads the published LisanBench snapshot with 5122.60 difficulty-weighted score. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on LisanBench?
The September 23, 2026 snapshot snapshot contains 154 AI models.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 2,000+ readers.
One email each week. Unsubscribe anytime.