# Capabilities and limits > Language models are fluent and broadly knowledgeable, and they fail in predictable ways. This topic names the failures a user must expect: confident invention, knowledge that stops at a date, answers that vary between runs, agreement that flatters instead of informs, instructions that fade as the context fills, and the cost and delay that grow with every token. Page: https://schubergphilis.github.io/ai-training/topics/concepts/limits/ License: CC BY-SA 4.0, https://creativecommons.org/licenses/by-sa/4.0/ *Concepts* ยท topic `concepts/limits` Language models are fluent and broadly knowledgeable, and they fail in predictable ways. This topic names the failures a user must expect: confident invention, knowledge that stops at a date, answers that vary between runs, agreement that flatters instead of informs, instructions that fade as the context fills, and the cost and delay that grow with every token. ## Concepts - **Hallucination**: Output that is fluent and confident but false: an invented citation, a function that does not exist, a plausible date that is wrong. It happens because the model predicts likely text, not verified facts, and it has no built-in signal for "I do not know". Hallucination is most likely on specifics, rare topics and anything the model cannot look up. [glossary](https://schubergphilis.github.io/ai-training/glossary/#hallucination) - **Knowledge cutoff**: The date after which a model's training data stops. Events, releases and documentation newer than the cutoff are unknown to the model unless they are pasted into the context or fetched by a tool. A model will often answer about recent things anyway, from older patterns, so the cutoff is a common and quiet source of wrong answers. [glossary](https://schubergphilis.github.io/ai-training/glossary/#knowledge-cutoff) - **Non-determinism**: The same prompt can produce different answers on different runs because output is sampled from probabilities. Even at the lowest temperature, small differences in serving can change results. This means one good answer does not prove a prompt is reliable, and any process built on a model needs checks that tolerate variation. [glossary](https://schubergphilis.github.io/ai-training/glossary/#non-determinism) - **Sycophancy**: The tendency of a model to agree with the user, praise their idea or change a correct answer when challenged, because agreeable text was rewarded during training. Sycophancy makes a model a poor judge of your own work unless you ask for criticism explicitly, hide your preference, or have it argue the other side. [glossary](https://schubergphilis.github.io/ai-training/glossary/#sycophancy) - **Cost and latency**: Every token read or written costs money and time. Long prompts, large documents, big models and multi-step agents multiply both. Latency grows with output length because tokens are produced one at a time. Cost and latency are why a smaller model or a shorter prompt is often the right engineering choice even when a larger one would answer slightly better. [glossary](https://schubergphilis.github.io/ai-training/glossary/#cost-and-latency) - **Instruction dilution**: The weakening of an instruction as more text piles up after it in the context window. A rule stated once at the start competes with everything said since; in a long conversation or a long instruction file the model follows some rules and quietly drops others. Dilution is why short instruction files work better than long ones and why a rule that must always hold belongs in a mechanical check, not in prose. [glossary](https://schubergphilis.github.io/ai-training/glossary/#instruction-dilution) - **Factuality and faithfulness**: Two kinds of hallucination that research on the subject separates. A factuality error is wrong about the world, such as an invented date or a citation that does not exist. A faithfulness error departs from what the person gave the model: it ignores an instruction, contradicts the document it was handed or contradicts itself. The first is found by checking an outside source, and the second by comparing the output with the input. [glossary](https://schubergphilis.github.io/ai-training/glossary/#factuality-and-faithfulness) ## Links - **Builds on:** [How language models work](https://schubergphilis.github.io/ai-training/topics/concepts/how-models-work/) - **Leads to:** [What an agent is](https://schubergphilis.github.io/ai-training/topics/concepts/what-is-an-agent/), [Recognizing failure](https://schubergphilis.github.io/ai-training/topics/safety/failure-modes/), [Responsible use](https://schubergphilis.github.io/ai-training/topics/safety/responsible-use/), [Choosing models and tools](https://schubergphilis.github.io/ai-training/topics/using-agents/choosing-tools/) - **Competencies drawing on it:** [Explains how a language model produces text and where it fails](https://schubergphilis.github.io/ai-training/competencies/concepts/explains-models/), [Recognizes an agent, its tools and its degree of autonomy](https://schubergphilis.github.io/ai-training/competencies/concepts/recognizes-agents/) ## Lessons - [Same prompt, different answer](https://schubergphilis.github.io/ai-training/concepts/same-prompt-twice/) (tutorial) - [Getting a straight answer](https://schubergphilis.github.io/ai-training/concepts/straight-answer/) (tutorial) - [Where a model makes things up](https://schubergphilis.github.io/ai-training/concepts/thin-path/) (explanation) - [What every token costs](https://schubergphilis.github.io/ai-training/concepts/what-tokens-cost/) (tutorial) - [Wrong about the world, or wrong about the text](https://schubergphilis.github.io/ai-training/concepts/wrong-about-the-world/) (explanation) ## Sources - `DLAI-3` [AI Prompting for Everyone](https://www.deeplearning.ai/courses/ai-prompting-for-everyone), DeepLearning.AI (course) - `Academy ai-capabilities-and-limitations` [AI capabilities and limitations](https://academy.claude.com/courses/ai-capabilities-and-limitations), Claude Academy (course) - `Academy ai-fluency-framework-foundations` [AI Fluency: Framework and foundations](https://academy.claude.com/courses/ai-fluency-framework-foundations), Claude Academy (course) - `Sharma 2023` [Towards Understanding Sycophancy in Language Models](https://openreview.net/forum?id=tvhaxkMKAn), International Conference on Learning Representations (ICLR 2024), arXiv preprint 2310.13548 (paper) - `Huang 2025` [A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions](https://doi.org/10.1145/3703155), ACM Transactions on Information Systems 43(2), Article 42, 2025, arXiv preprint 2311.05232 (paper)