Independent benchmark desk / September 19, 2026 baseline
Claude Reports
Source-first analysis of Claude, GPT, and agentic AI benchmarks. As of September 19, 2026 the current Claude lineup is Fable 5.1, Opus 5, Sonnet 5, and Haiku 4.5. We preserve the part most leaderboard screenshots omit: model access, harness, effort, cost, contamination risk, and what the score is allowed to prove.
Signature experience
The Journal
A bound reading journal for the September 2026 rate card and one sourced table: Fable 5.1 versus Opus 5, using only Anthropic’s public announcement, labeled vendor numbers. Open it. Turn the pages. Margin notes carry harness, effort, contamination, and the $0.25 cache-read exception.
Open the JournalVol. I · September 2026
The Journal
Fable 5.1 is $10/$50. Opus 5 is $5/$25. Sonnet 5 stayed $2/$10. The scores inside are Anthropic-reported, not independent measurements.
Featured analysis
The lineup changed twice. The caveats did not.
Opus 5 landed in July and Fable 5.1 on September 1. Anthropic published a comparison table with both — provider-run, scaffold undisclosed, with GPT-5.6 Sol only where they published a figure. We reproduce it and say what it can and cannot prove.
Model Lineup
Claude Fable 5.1 vs Opus 5 vs Sonnet 5: Reading Anthropic's September 2026 Lineup
A source-first read of the current Claude lineup after the September 1, 2026 Fable 5.1 launch, including Anthropic's own benchmark table and the harness caveats that table does not carry.
Read the report
Cost
Sonnet 5 Pricing
Sonnet 5 at $2/$10 is a standing price, not a countdown. Any routing decision that was deferred because a 50% input increase looked imminent should be re-run against the current pricing page.
Policy
Text Watermarking
Watermarking is a provenance signal about generation, not an identity or attribution system. Do not build authorship enforcement, plagiarism adjudication, or user-level tracing on top of it.
Safety
Eval Incident Disclosure
Eval containment is part of benchmark validity, not a separate safety topic. A lab that publishes containment failures and submits to third-party review is handing you information that a clean results table structurally cannot contain.
Free tools
Turn claims into working estimates.
The tools are client-side and source-linked: model comparison, benchmark decoding, and Claude-vs-OpenAI API cost estimation, all carrying the current Fable 5.1, Opus 5, Sonnet 5, and Haiku 4.5 rows.
Tool
Model comparison
Filter Claude, GPT, and benchmark-reference models by provider, access, price, context window, and published coding benchmark evidence.
Tool
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.
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.
Primary-source posture
"SWE-Bench Verified is increasingly contaminated."
OpenAI's warning is one reason this site reports benchmark provenance, not just winner labels.Read the source.
Keep reading
Related reports and tools
Internal links are selected to keep benchmark, tool, and methodology context together.
Journal
The Journal
Bound reading journal for September 2026 cost per MTok and Anthropic-reported Fable 5.1 vs Opus 5 scores.
Report
Fable 5.1 vs Opus 5 vs Sonnet 5
Start with Opus 5, escalate to Fable 5.1 when your own evals at higher effort still fall short, and route volume to Sonnet 5. Treat Anthropic's launch table as a directional provider claim, not an independent bake-off.
Report
Sonnet 5 Pricing
Sonnet 5 at $2/$10 is a standing price, not a countdown. Any routing decision that was deferred because a 50% input increase looked imminent should be re-run against the current pricing page.
Tool
Free benchmark tools
Use client-side calculators and explainers for model comparison, benchmark reading, and API cost planning.