Skip to main content
Radar

Five or fewer confirmed AI changes, with original sources, on mornings when something changed.A free source-linked morning brief.

Start free brief
BenchLM

TruthfulQA

A benchmark designed to measure whether language models produce truthful answers instead of repeating common misconceptions or misleading falsehoods.

Data verified 25 confirmed releases in the last 30 daysStart free brief

How BenchLM shows TruthfulQA right now

BenchLM is tracking TruthfulQA in the local dataset, but exact-source verification records for these rows are still being attached. To avoid a blank benchmark page, BenchLM shows the current tracked rows below as a display-only reference table.

These tracked rows are useful for inspection and spot-checking, but until exact-source attachments are completed they should not be treated as fully verified public benchmark rows.

Snapshot

2 tracked modelsLocal tracked rowsAwaiting exact-source attachmentsDisplay only
TruthfulQA: Measuring How Models Mimic Human Falsehoods

Tracked score on TruthfulQA — August 21, 2026

We mirror the published tracked score view for TruthfulQA. Phi-4 leads the public snapshot at 77.5%, followed by Kimi K2.5 (57.3%). We do not use these results to rank models overall.

1Open

Phi-4

Microsoft

77.5%
Overall 22.42Context 16K
2Open

Kimi K2.5

Moonshot AI

57.3%
Overall 58.68Context 256K
2 modelsKnowledgeStaleDisplay onlyUpdated August 21, 2026

Tracked score table (2 models)

Score
1
Phi-4Microsoft · Open weight
77.5%
2
Kimi K2.5Moonshot AI · Open weight
57.3%

About TruthfulQA

Year

2021

Tasks

Truthfulness and misconception resistance

Format

Question answering

Difficulty

Hallucination and factuality stress test

TruthfulQA matters because many models sound confident while repeating popular but false answers. It is a useful factuality and hallucination-adjacent benchmark even though it is older than newer factuality suites.

TruthfulQA: Measuring How Models Mimic Human FalsehoodsPublic benchmark source

BenchLM freshness & provenance

Version

TruthfulQA 2021

Refresh cadence

Static

Staleness state

Stale

Question availability

Public benchmark set

StaleDisplay only

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.

FAQ

What does TruthfulQA measure?

A benchmark designed to measure whether language models produce truthful answers instead of repeating common misconceptions or misleading falsehoods.

Which model leads the published TruthfulQA snapshot?

Phi-4 currently leads the published TruthfulQA snapshot with 77.5% tracked score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on TruthfulQA?

2 AI models are included in BenchLM's mirrored TruthfulQA snapshot, based on the public leaderboard captured on August 21, 2026.

Last updated: August 21, 2026 · mirrored from the public benchmark leaderboard

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 2,000+ readers.

One email each week. Unsubscribe anytime.