Particle AI: Particle-1

Particle-1 from Particle AI delivers strong overall capabilities, pairing top-tier execution speeds with competitive pricing across comprehensive evaluations. Operating in the 73rd percentile for speed while maintaining balanced operational costs, the model also demonstrates exceptional technical reliability, recording a perfect 100% success rate with zero system or API failures. Performance across benchmark categories is anchored by exceptional analytical and technical strengths. Most notably, Particle-1 achieved a perfect 100% accuracy in complex reasoning, distinguishing itself as the most accurate model at its price point and within its speed class. It also excels in programming, securing a 95.0% accuracy score (87th percentile) in coding assessments, alongside strong mathematical proficiency at 94.0%. The model's primary relative weakness appears in instruction following, where it attained a lower 78.0% accuracy when navigating complex, multi-layered constraints. Nevertheless, bolstered by its massive 1M-token context window and specialized post-training, Particle-1 offers an outstanding balance of speed, affordability, and elite problem-solving, making it particularly well-suited for demanding software engineering tasks and data-heavy agentic workflows.