Loading model…
deepseek/deepseek-v4-flash14/27 v1 — Nutrition · score 48.3% · pass rate 48.5% (181/373); 1/2 v2 — Resilience · score 71.2% · pass rate 49.2% (226/459)2.21s + output speed 1006 tok/s$0.09, prompt tokens 47332, completion tokens 299320$0.09 across 832 tests with high output speed (1006 tok/s).1/2), outperforming grok-4.3 by 3.6 score points.cost category (29/30 passed) and speed (85/125 passed).31/122 passed) and refactoring (35/120 passed) scores drag down overall results.coding_ui (0/6), indicating UI generation is a significant gap.14/27) with sub-50% pass rate suggests struggles with foundational tasks.grok-4.3 by 37.1 score points and 37.5 pass rate points.grok-4.3 by `Overall score % from merged run_models rows (chronological). Only runs that include this model appear as points.
Overall score % from the most recent run per selected suite version for this model. Different suites use different fixtures and max scores — only interpret comparisons qualitatively.
No score samples for the selected versions.
Score % vs pass rate % per category. With 0/1 scorers, both usually line up; with proportional tests, score % reflects partial credit while pass rate counts tests that clear the fixture threshold.
Total estimated spend per scope for this model (bars, left axis) and mean spend per merged result row (line, right axis: total ÷ tests).
Pass rate % per difficulty level — complements the score % view above.
Normalized 0–100 within this model: TTFT (shorter → higher spoke) and decode tok/s (higher → higher spoke). Values come from streamed BLXBench runs merged into overall_ranking.json.
Pass rate % per category for this model (distinct from score %, which reflects partial credit).