The TKX

The benchmark for the cost of AI capability. One fixed scoring recipe, tracked over time, measuring what a unit of standardized intelligence costs across open and closed models.

Cite the print: GET /v1/index/tkx · Methodology

Price this against your stack →

Each index is set to 100 at the 2026-08-04 baseline and moves as prices and quality on the current members change. Above 100 means a unit of AI capability is at a premium to that baseline. Below 100 means it is lower than at that baseline. Composite quality follows a fixed standard workload: 30% reasoning, 30% code, 25% extraction, 15% classification. The headline basket holds general-purpose models scored across all four families; specialists rank in their task families only. When membership changes, the published reading is chained so the print does not jump; the composition effect is shown on those days only. Version ledger: https://cambent.ai/methodology

Model value is task-conditional: quality is scored per task family, and a model ranks only in families it serves. Pick a task and watch the field change — specialists enter where they compete and are absent where they don't.

Every quality score is labeled Grounded, Partially grounded, or Provisional. Grounded means published benchmarks and a sourced price for that exact model. We show the source and date. Partially grounded means some scores for this exact model are from a named source; others are not published yet. We say which. Missing stay unlabeled. Never estimated from a close version. Provisional for list-only means listed, not scored yet. Other provisional estimates say grounding is in progress without inventing a score. Today: 11 Grounded · 6 Partially grounded · 19 Provisional of 36 models.

Scores are labeled by how we know them. Details → https://cambent.ai/methodology

Ranked 24 of 36 models for this task · specialists rank only in their families

24 live prices with a known source, 12 manual. Today: 11 Grounded · 6 Partially grounded · 19 Provisional of 36 models. Specialists appear only in their own task lists.

DeepSeek V4 Flash Provisional listed, not scored yet. Qwen 3.7 Plus Provisional listed, not scored yet. Mistral Medium 3.5 Provisional listed, not scored yet. Mistral Small 4 Provisional listed, not scored yet. Codestral Provisional listed, not scored yet. GPT-6 Astra Provisional listed, not scored yet. Claude Fable 5.1 Provisional listed, not scored yet. Llama 4 Maverick Provisional listed, not scored yet. DeepSeek V4.1 Flash Provisional listed, not scored yet.

Muse Spark 1.3 Provisional listed, not scored yet. Qwen 3.8 Max 0902 Provisional listed, not scored yet.