Value Density Score
What is value density score?
Value Density Score measures benchmark performance per dollar — how many MMLU points you get for each dollar of input cost per million tokens. A model with 85% MMLU at $3.00/1M has a VDS of 28.3. The same capability at $0.15/1M has a VDS of 566.7. This is the metric for answering: which model gives me the most reasoning capability per dollar?
Why it matters
VDS answers: if I have $1 to spend on model API calls, which model gives me the most reasoning capability? At the cheap end, Gemini 1.5 Flash delivers roughly 1,133 MMLU points per dollar. At the expensive end, GPT-4 delivers about 2.9. That is a 390× efficiency gap — and it changes every time a provider adjusts pricing. sourc.dev computes VDS automatically when either benchmark or pricing data changes.
Where models stand
Data available for 24 of 271 tracked entities. Last updated 2026-03-31.
How sourc.dev tracks this
sourc.dev verifies value density score manually from official provider documentation, API responses, and published specifications. Every data point includes a source URL and verification date. When a value changes, the old value is preserved in the history table and the new value is recorded alongside it. Nothing is overwritten — the full timeline is always available.
sourc.dev verifies this attribute manually from provider documentation. Every data point includes a source URL and verification date. Changes are recorded in the history table — nothing is overwritten.
This attribute is verified periodically against provider documentation. When sourc.dev detects a change, the new value is recorded alongside the old one with full provenance.
Understanding value density score helps developers make informed decisions when choosing between models and providers. Rather than relying on marketing claims, sourc.dev provides verified, dated, source-linked data so the data decides.