Access benchmark data the way your workflow needs

We run a full evaluation sweep weekly, every Tuesday. When a model provider releases a significant update, we schedule an out-of-cycle run within 72 hours. The last-updated date on each table row shows exactly when that model's scores were last computed.

Yes. Pro accounts can submit up to 3 models per month for private evaluation runs. Results are visible only to your account. Enterprise accounts have no submission limit and can run models against custom task suites. Contact us for details on the submission process.

Code generation is tested across 400+ programming tasks covering algorithmic problems, API usage, and debugging. Image quality uses a composite score of perceptual quality (FID), diversity index, and prompt adherence rate. Video coherence evaluates temporal consistency and motion quality across 200+ prompts. Full methodology details are on the Benchmarks page.

Yes. Switching the billing toggle above to Annual gives you 20% off on Pro and Enterprise tiers. That works out to $119/month for Pro (vs $149/month) and $639/month for Enterprise (vs $799/month), billed as a single annual payment.

The API is a REST JSON API. After creating an account, you get an API key from your dashboard. You can query current leaderboard rankings, historical score time series for individual models, and category-level breakdowns. Free tier: 100 calls/day. Pro: 10,000 calls/day. All responses include a last_updated timestamp and confidence interval metadata.