Particle-1 cuts coding-agent tokens nearly in half

Token efficiency is becoming a model capability in its own right, and Particle-1 makes a credible case that shorter agent trajectories can reduce cost without sacrificing results.

• –Evaluated on Terminal-Bench 2.1, ReactBench, ADE Bench, and DeepSWE 1.1
• –Particle reports 34.4–51.6% lower average task costs than DeepSeek V4.1 Flash
• –The savings come from reduced reasoning and tool-call tokens, which could make long-running coding agents more practical
• –Results are first-party claims and still need independent reproduction across real-world codebases
• –OpenRouter pricing is $0.25 per million input tokens and $1 per million output tokens, with a 1M-token context window