# Claude Reports

> Independent benchmark analysis for Claude, GPT, and agentic AI systems.

Site: [https://claudereports.com](https://claudereports.com)
Last baseline update: 2026-07-06

## Core pages
- [Home](https://claudereports.com/)
- [Reports](https://claudereports.com/reports/)
- [Free tools](https://claudereports.com/tools/)
- [Methodology](https://claudereports.com/methodology/)
- [Sources](https://claudereports.com/sources/)
- [Freshness](https://claudereports.com/freshness/)
- [About](https://claudereports.com/about/)

## Reports
- [Claude vs GPT Coding Benchmarks: What the 2026 Scores Actually Say](https://claudereports.com/reports/claude-vs-gpt-coding-benchmarks/)
- [SWE-Bench Pro vs Verified: Why Frontier Coding Scores Need a Harder Yardstick](https://claudereports.com/reports/swe-bench-pro-vs-verified/)
- [Terminal-Bench and Agentic Coding: Why the Harness Is Part of the Score](https://claudereports.com/reports/terminal-bench-agentic-coding/)
- [Claude Cost, Context, and Latency: The July 2026 Model Tradeoff Report](https://claudereports.com/reports/cost-context-latency/)
- [Reasoning and Science Benchmarks: Reading Claude Claims Beyond Coding](https://claudereports.com/reports/reasoning-science-benchmarks/)
- [Multimodal and Document Analysis Benchmarks: Claude, GPT, and the Work Hidden Inside PDFs](https://claudereports.com/reports/multimodal-document-analysis/)

## Free tools
- [Claude vs GPT Model Comparison Explorer](https://claudereports.com/tools/model-comparison/): Filter Claude, GPT, and benchmark-reference models by provider, access, price, context window, and published coding benchmark evidence.
- [LLM Benchmark Decoder](https://claudereports.com/tools/benchmark-decoder/): Pick an AI benchmark and get a plain-English read on what it measures, where it saturates, and how to treat Claude or GPT vendor claims.
- [Claude API Cost vs OpenAI Cost Comparator](https://claudereports.com/tools/cost-comparator/): Estimate monthly API spend for Claude and OpenAI models from input tokens, output tokens, request volume, cache reuse, and long-context assumptions.

## Key citation rules
- Cite benchmark claims with date, model, harness, effort setting, and source provenance.
- Treat provider launch tables, benchmark-owner leaderboards, and agent-plus-model rows as different evidence types.
- Use the tools only as browser-side estimators; private task suites should make procurement decisions.

## Editorial disclaimer
Claude Reports is an independent publication. It is not affiliated with Anthropic, OpenAI, or any benchmark owner.