Particle-1 launches on OpenRouter with 1,049k context at $0.25/M input

What changed

Particle AI has listed Particle-1 on OpenRouter, a token-efficient agentic coding model post-trained on DeepSeek V4.1 Flash. It is positioned for software engineering work including terminal workflows, codebase changes, and data tasks. The listing carries a 1,049k token context window and public per-token pricing.

Specs

• •Context window 1,049k tokens
• •Pricing input $0.25 per M tokens
• •Pricing output $1.00 per M tokens
• •Model ID particle-ai/particle-1
• •Vendor docs https://openrouter.ai/particle-ai/particle-1

Why it matters

At $0.25 per M input and $1.00 per M output, Particle-1 undercuts Claude Sonnet 4.5 at $3 per M input and $15 per M output by more than an order of magnitude on both sides. The 1,049k context window means a Vibe Builder can hand it a whole small repo or a long terminal session log in one pass instead of chunking files. Developers get a cheap agentic coding option to route against DeepSeek V4.1 Flash and Qwen3-Coder when cost per task matters more than peak reasoning.

What to watch for

Compare Particle-1 against DeepSeek V4.1 Flash and Qwen3-Coder on the same agentic coding tasks before switching any production route. Run your own terminal and codebase-edit evals, since post-training on a Flash base can shift behaviour on long tool-call chains. Watch output token counts per task, because a cheap input rate still adds up if the model is verbose in agent loops.