INDEPENDENT AI MODEL BENCHMARK

QBench Index 1.6

A source-audited view of model capability. Compare exact effort variants, inspect benchmark evidence and explore the trade-offs between capability, coverage and price.

SCORING FRAMEWORK

08 capability familiesEqual-weight observed family means
Exact reported effort variants
Missing scores are never imputed

QBench Index 1.6 leaderboard

How QBench Index 1.6 works

Each benchmark keeps its reported 0–100 accuracy or task-success percentage. Results are averaged inside eight capability families, then the observed family means are averaged equally. Missing families are left out, not imputed and not pulled toward 50. Model effort variants are separate entries; unreported effort is never guessed.

Rank: at least 3 eligible evaluations across 3+ capability families · Evidence: full (all 8 families), broad (4–7), or limited (3) · Insufficient evidence: retained in Models, without an Index score.

VISUAL ANALYSIS

Benchmark intelligence workspace

Explore score composition, model rankings, benchmark coverage and price trade-offs.

SOURCE-BASED DATA

MODEL EXPLORER

Model catalog

Explore measured model and effort variants, compare evidence and inspect official API pricing and inventory without mixing unranked models into the Index.

MODEL DATA

Measured + cataloguedExact model/effort records
Source-separated API pricing

Models

Models with at least one eligible result. Routers and specialist systems are identified separately. The API price catalog below also includes priced models outside the composite Index.

API price catalog

Per 1M input/output tokens. Official provider API and OpenRouter route prices are labelled separately.

Model	Provider	Input / 1M	Output / 1M	Price source	Conditions

Verified API inventory — not yet ranked

Officially documented model IDs that do not have an exact eligible QBench benchmark result. These models are not assigned scores or leaderboard positions.

Model	Provider	Official API ID	Context	API price	Status

Model comparison

Exact model/effort variants · missing evidence stays blank · unranked variants remain visibly unranked

Comparison rules: Index values are shown only for variants that pass QBench Index 1.6 eligibility. Benchmark rows retain their publisher-reported scores and effort labels; scores from different benchmarks are not directly interchangeable. A dash means no eligible reported result, not zero.

PROVIDER LANDSCAPE

Provider analysis

Compare provider coverage, measured variants, ranked evidence and price-catalog depth. Aggregate scores are descriptive, not provider quality ratings.

COMPARISON BASIS

Measured variantsRanked evidence separated
Aliases normalized transparently

Provider landscape

Provider coverage from source-audited eligible results. Scores include ranked variants only.

The chart shows the mean QBench Index of ranked variants for the 15 providers with the most ranked variants. The table includes every provider represented in the eligible result set. It is not a provider quality rating.

Provider	Model names	Variants	Ranked	Mean Index	Priced models

REAL-WORLD TASK ECONOMICS

Cost & efficiency

Inspect reported task cost, duration and token usage within each benchmark and harness. Different evaluations are never pooled into misleading averages.

MEASUREMENT RULE

Reported data onlyNo estimated task costs
Harness context stays visible

Measured task economics

Compare cost, duration or token usage within one benchmark and its stated harness. Metrics from different benchmarks are not pooled.

Only publisher-reported measurements are shown. Missing cost, duration and token fields are not estimated.

Model	Provider	Effort	Score	Cost / test	Cost / successful task	Duration / test	Tokens / test	Source

EVALUATION LIBRARY

Benchmark directory

Browse source-audited evaluations, versions, snapshots and model-level results. Select any benchmark to inspect its underlying evidence.

EVIDENCE FIRST

Publisher-level resultsVersion and snapshot metadata
Eligibility shown explicitly

Benchmark directory

Select a benchmark to see its model scores. The status shows whether it enters the index.

Select a benchmark

TRANSPARENT EVALUATION

Methodology & evidence

Understand how QBench selects evaluations, handles missing results, computes the composite and assigns evidence tiers.

CURRENT FRAMEWORK

QBench Index 1.6Equal-weight family means
Source-audited benchmark inputs