Gemini 4 hallucinates only ~4% of the time, GPT-6.1 Terra hits 34.5%
shensi · x · 2026-11-10
Merge published a hallucination-rate comparison of top LLMs based on Artificial Analysis data (week of Nov 9, 2026).
- Gemini 4 Carbon (medium) at 4.2% and Gemini 4 Argon (high) at 4.3% lead by a wide margin, making up an answer only 4% of the time when they don't know one
- Meta's Muse Wonder Full (max) is second at 17.6%
- GPT-6.1 Terra (max) hits 34.5% and Kimi K3.1 Pro 37.3% — roughly as many correct answers as Gemini 4, but nearly 10x the fabrication rate
Merge uses the data to promote its gateway, which routes each request to the best-fitting model without changing integrations.
More from Models
- NeuroscienceBench debuts: 91 evals test if AI agents can analyze real neuroscience data — kenbwork · 2026-11-10
- Gemini 4 Carbon beats Claude Opus 5.5: 23% lower cost per task, #1 on Intelligence Index — DeepLearningAI · 2026-11-10
- Gensyn live-trains open-8b: every step auditable on your laptop, overhead down to 1.3x — benfielding · 2026-11-10
- Meta's Muse Ember matches Claude Haiku 5.5 on Intelligence Index at a third of the cost — ArtificialAnlys · 2026-11-10
- Scale AI launches Muse Ember, a small model claiming Haiku-level intelligence at 1/3 the cost — alexandr_wang · 2026-11-10
- Opus mocked for claiming everything's been done before, then flip-flopping after a web search — teortaxesTex · 2026-11-10