Press & research
Citable statistics and data exports
Anthropic: Claude Opus 5.5 (batch) leads on real Artificial Analysis quality benchmarks at an Intelligence Index of 59.1 as of November 7, 2026, among 96 models with published benchmark scores.
Everything below is derived from OpenRouter's and Artificial Analysis's public data and computed daily. Cite any figure with the date shown; every number here is independently checkable against the published methodology.
Today's citable statistics
- 01Anthropic: Claude Opus 5.5 (batch) leads on real Artificial Analysis quality benchmarks at an Intelligence Index of 59.1 as of November 7, 2026, among 96 models with published benchmark scores.
- 02Ox Alpha handles the largest real share of production traffic on OpenRouter at 27.4% of tracked token volume, averaged over the trailing 125 days.
- 03OpenAI: gpt-oss-120b (batch) delivers the best measured quality-per-dollar among tracked models, at 692.5 intelligence-index points per dollar.
- 04Open-weight models handle 78% of real token volume on OpenRouter as of November 6, 2026.
- 05404 distinct AI models are tracked daily, spanning 72 providers.
- 06Real production usage is classified across 29 distinct task categories (from memory extraction to translation), each measured directly from OpenRouter's classified traffic - not survey or benchmark opinion.
Source: OpenRouter (openrouter.ai/rankings), as of November 7, 2026.
Benchmark scores: Artificial Analysis (artificialanalysis.ai) via OpenRouter (openrouter.ai/rankings).
Methodology: www.smophy.ai/benchmark/methodology
Token counts originate from each provider's own tokenizer and are not directly comparable across providers.
Data exports
Real derived metrics only - quality-per-dollar and usage share, tracked daily. Raw OpenRouter rows are never re-served, per their terms (see attribution below).
One-paragraph methodology
No invented composite index. Every published number is either a real, independently sourced quantity (benchmark score, price, usage share, uptime) or a plain, transparent ratio of two such numbers (quality ÷ price). Full detail, including the history of three composite indices we built and killed after finding real defensibility problems, is on the methodology page.
Attribution & contact
When citing these figures, please link to the relevant page on www.smophy.ai/benchmark and note the date shown. Underlying usage and pricing data originates from OpenRouter (openrouter.ai/rankings); benchmark scores from Artificial Analysis (artificialanalysis.ai) via OpenRouter.
Our derived statistics (quality-per-dollar, usage share, and every other computed number on this site) are published under CC BY 4.0 - free to reuse with attribution.
SmophyAI Field Tests: citation and raw data
First-party real-work tests run by Kamil Kępiński, founder of SmophyAI. Every published test provides the frozen prompt, verifiable checks and free raw answers.
0 published tests · 0 tests with no material difference · 0 model versions on the board.
Field Test Board and CSV/JSON downloads · Method v1
Suggested citation
SmophyAI Field Tests, run by Kamil Kępiński, <test id>, <date>, www.smophy.ai/benchmark/field-tests/<test-id>