Source index

Primary Sources

The source desk behind Claude Reports. These are the official docs, release notes, system-card indexes, and benchmark-owner pages used for the September 19, 2026 baseline.

Citable baseline

Source index facts

Dated facts with primary-source links where the underlying source is available.

23Primary sources indexed for reports, tools, and llms-full coverage.
11Anthropic source records for Claude model specs, pricing, launches, and system cards.
4Benchmark-owner source records for coding, terminal-agent, and reasoning context.

Anthropic

Anthropic Claude Platform Docs

Models overview

Current Claude model IDs, pricing, context windows, effort defaults, knowledge cutoffs, and retirement dates for Fable 5.1, Opus 5, Sonnet 5, and Haiku 4.5.

Anthropic

Claude Fable 5.1 and Claude Mythos 5.1

The September 1, 2026 launch announcement. Contains the provider-run benchmark table comparing Fable 5.1, Fable 5, Opus 5, and, where Anthropic published a figure, GPT-5.6 Sol, plus cache-read pricing and safeguards changes. Those Sol figures are Anthropic-reported, not OpenAI-reported and not independent.

Anthropic Claude Platform Docs

Claude Fable 5.1 overview

API-facing Fable 5.1 behavior: effort defaults, thinking-block handling, breaking changes against Fable 5, and additive betas.

Anthropic

Claude Opus 5

Opus 5 launch on July 24, 2026: pricing held at Opus 4.8 rates, fast mode, plan availability, and safety-classifier fallback routing.

Anthropic Claude Platform Docs

Claude Opus 5 overview

API-facing Opus 5 context window, max output, adaptive thinking, effort defaults, and platform betas.

Anthropic

Introducing Claude Sonnet 5

Sonnet 5 launch, cost-performance positioning, and the original introductory-pricing window. Read alongside the current pricing page, which now lists $2/$10 as standard.

Anthropic

Text watermarking in Claude

What Claude text watermarking is and is not: EU AI Act and Code of Practice context, no hidden characters, no added tokens, no person-level traceability, detection API forthcoming.

Anthropic Claude Platform Docs

Choosing the right model

Anthropic model-selection guidance, including the recommendation to start with Opus 5 and escalate to Fable 5.1 for demanding reasoning and long-horizon agentic work.

Anthropic Claude Platform Docs

Pricing

Claude per-token pricing, cache-read and cache-write multipliers, batch discount, and the standing $2/$10 Sonnet 5 rate.

Anthropic

Model system cards

Index of system cards for current and earlier Claude models, including the Fable/Mythos and Opus lines.

OpenAI

OpenAI

Introducing GPT-5.5

GPT-5.5 benchmark claims, including SWE-Bench Pro and Terminal-Bench 2.0.

OpenAI API Docs

GPT-5.5 model docs

GPT-5.5 API pricing, context window, output limit, modalities, and model snapshot information.

OpenAI API Docs

GPT-5.4 model docs

GPT-5.4 API pricing, context window, output limit, modalities, and long-context pricing caveats.

Benchmark

ARC Prize

ARC-AGI Series

ARC-AGI benchmark family and its focus on skill acquisition, abstraction, and generalization.

Research

Keep reading

Related verification surfaces

Internal links are selected to keep benchmark, tool, and methodology context together.

Method

Benchmark methodology

See how Claude Reports reads provenance, harness, effort, contamination, uncertainty, and cost.

Tool

Model comparison

Filter Claude, GPT, and benchmark-reference models by provider, access, price, context window, and published coding benchmark evidence.

Tool

Cost comparator

Estimate monthly API spend for Claude and OpenAI models from input tokens, output tokens, request volume, cache reuse, and long-context assumptions.

Freshness

Freshness log

Track model launches, benchmark revisions, source changes, and future weekly updates.