Model Class

General-Purpose LLM

A large language model for broad text tasks. In scope when applied to metadata extraction, summarization, schema inference, or querying of S3-stored content.

10 connections 3 resources

Summary

What it is

A large language model for broad text tasks. In scope when applied to metadata extraction, summarization, schema inference, or querying of S3-stored content.

Where it fits

General-purpose LLMs are the most versatile tool in the LLM-Assisted Data Systems topic. They can extract metadata, infer schemas, classify documents, and generate SQL — all tasks that previously required custom engineering for each S3 dataset.

Misconceptions / Traps
  • General-purpose LLMs are not deterministic. The same prompt can produce different outputs. For production pipelines, use structured output constraints and validation.
  • Context window limits constrain how much S3 data can be processed per call. Large documents or schemas may need chunking strategies.
Key Connections
  • enables Metadata Extraction, Schema Inference, Natural Language Querying, Data Classification — the model class behind all four capabilities
  • Code-Focused LLM is_a General-Purpose LLM — a specialization
  • scoped_to LLM-Assisted Data Systems

Definition

What it is

A large language model trained on broad text data, capable of understanding and generating natural language across many domains.

Why it exists

General-purpose LLMs can interpret the content of S3-stored objects — extracting metadata, inferring schemas, classifying documents, and translating natural language to SQL — tasks that previously required manual engineering or domain-specific tools.

Primary use cases

Metadata extraction from S3-stored documents, schema inference over semi-structured S3 data, natural language querying of S3-backed datasets.

Recent developments

Latest signals
  • 2026 frontier landscape: GPT-5 family + Claude Opus + Gemini 3 Pro + Grok 4 + DeepSeek V3/R1 + GLM-5. Leading general-purpose LLMs in 2026 are Claude Mythos Preview, Claude Opus 4.6, the GPT-5 family, Claude Opus/Sonnet, Gemini 3 Pro, xAI Grok 4, DeepSeek V3/R1, and Z.AI GLM-5. Per LLM Stats — LLM Leaderboard 2026.
  • GPQA Diamond leaders: Claude Mythos Preview at 94.6%, Grok 4 at ~87-88%, Gemini 2.5 Pro near 86%, GPT-5 at 83-85%. Most discriminating reasoning benchmark at the frontier. Per LM Council — AI Model Benchmarks May 2026.
  • Pricing band: $0.10/$0.40 (Gemini 3.1 Flash-Lite) → $30/$180 (GPT-5.4 Pro) per M tokens. Frontier API pricing spans roughly 3 orders of magnitude. GPT-5.4 = $2.50/$15.00 per M; Claude Sonnet = $3.00/$15.00; Haiku $0.25/$1.25; DeepSeek V4 = $0.435/$0.87 per M (5-70× cheaper than US frontier). Per TechBullion — 2026 LLM API Pricing Comparison.
  • DeepSeek V4 = 5-70× cheaper than US frontier models on equivalent benchmarks. Pricing pressure from open-weight Chinese frontier is the dominant 2026 market force. Per TechBullion — 2026 LLM API Pricing.
  • SWE-bench Verified at 72-80% for top models. Claude Opus 4.7 in 70-80% range, Grok 4 around 75% on autonomous resolution, GPT-5 near 75% on Verified runs. Per Vellum — LLM Leaderboard 2026.
  • 2026 LLM comparison guides aggregating 300+ models. LLM Stats's 2026 leaderboard tracks 300+ top AI models by intelligence + speed + price — vendor-neutral procurement reference. Per LLM Stats — Leaderboard 2026.

Connections 10

Outbound 5
Inbound 5

Resources 3