Atlas

Benchmarks

← All benchmarks

BLXBench Hallucination category

Code · 2026-04-25

Hallucination category component of the BLXBench public leaderboard. The source describes this category as grounded answers under adversarial or missing-context prompts; scores are category score percentages.

Top models (higher is better)

ModelScore
Opus 4.884.2
MiniMax M380.8
Claude Fable 579.7
GPT-5.578.9
GLM-5.277.5
Opus 4.777.2
Kimi K2.676.9
MiMo-V2.5-Pro76.4
Nemotron 3 Super 120B A12B76.1
DeepSeek-V4-Pro75.8
Step 3.7 Flash75.8
Kimi K2.7 Code74.7
MiMo-V2.574.7
GPT-5.3-Codex74.4
Qwen3.7-Max74.4
Ring 2.6 1T74.1
North Mini Code73.9
GLM-5.173.3
Granite 4.1 8B73.3
Qwen3.7-Plus73.3
Qwen3.6-Flash-2026-04-1672.8
CoBuddy72.5
DeepSeek-V4-Flash72.5
Nemotron 3 Nano 30B A3B71.4
Grok Build 0.170.8
Grok 4.370.6
Mistral Medium 3.570.6
Gemini 3.1 Flash-Lite66.9
Mistral Small 465.0
Nemotron 3 Nano Omni 30B A3B60.8
MiniMax M2.756.1
Gemini 3.5 Flash31.9
Loading Atlas data…