Seed 2.0 Pro (Non-reasoning): benchmark results

Provider: ByteDance. Released 2026-02-14. Access: API.

Unified ELO 1606 ± 1, rank #411 of 1935 rated models, from 27 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
RuleWeaver - Cross-Source - Rule Precision75.73Correctly applied share of cited rules (%)90
EgoBench - Dynamic Hard Mode - Micro Accuracy37.22MicroAcc (%; share of ground-truth tool calls issued, pooled85.7
EgoBench - Static Mode - Micro Accuracy40.7MicroAcc (%; share of ground-truth tool calls issued, pooled85.7
VitaBench 2.042.8Avg@4 Full Context (self-reported)75
EgoBench - Dynamic Easy Mode - Micro Accuracy38.69MicroAcc (%; share of ground-truth tool calls issued, pooled71.4
EgoBench - Static Mode13.3Joint Success Rate (%; every required tool call issued and t71.4
ComboShoppingBench - Response Quality88.7Pass rate (%; LLM-judged)59.5
EgoBench - Dynamic Easy Mode14.74Joint Success Rate (%; every required tool call issued and t57.1
EgoBench - Dynamic Hard Mode11.58Joint Success Rate (%; every required tool call issued and t57.1
RuleWeaver - Cross-Source - Rubric Score33.02Judge rubric score (0-100)50
RuleWeaver - Same-Source - Rule Precision61.63Correctly applied share of cited rules (%)40
LLMEval-Logic Formalization Free30.9Accuracy (%)38.5

Interactive version: theaggregate.ai/model?slug=seed-2-0-pro-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-25.