Particle AI · Available now

Particle-1 by Particle AI (2026): price per reply, price per million tokens and context window

Particle-1 is Particle AI’s model, post-trained on DeepSeek V4.1 Flash and launched on October 28, 2026, with a 1,048,576-token context window and a peak price of $0.25 per million input and $1.00 per million output tokens on Particle’s own API.

Available nowLast updated:

Use it now

Chat

≈ 1 credit per reply · See all models

New accounts get 5 free credits once, valid for 7 days. They work on the lower-priced models here; the priciest ones need credits you bought.

Start chatting — 5 free credits
Maker
Particle AI
Type
Text
Released
2026-10-28
OpenRouter id
particle-ai/particle-1

Per million tokens

Price

Particle-1 · checked 2026-11-17
-InputOutputSource
Particle AI — Peak (standard)$0.25$1docs.particle.ai
Particle AI — Off-peak$0.13$0.50docs.particle.ai
Via OpenRouter (particle-ai/particle-1)$0.06$0.50openrouter.ai

Checked

Facts, with sources

  • Particle AI’s launch post, dated October 28, 2026, describes Particle-1 as "post-trained on DeepSeek V4.1 Flash". Particle says it trained the model with reinforcement learning in its own environments to solve tasks with fewer tokens while keeping accuracy, using a rank-64 LoRA adapter whose trainable parameters are about 1.25% of the model’s parameter count. It says the first release focuses on software engineering. Particle’s API docs call it "Particle’s own frontier model, recommended for chat, agentic coding, and tool-calling workloads".Source: particle.ai · checked 2026-11-17
  • The maker’s own figures, not tested by us: Particle says Particle-1 uses "around ~45% fewer tokens on average" than DeepSeek V4.1 Flash. Across four coding and data benchmarks (Terminal-Bench 2.1, ReactBench, ADE Bench, DeepSWE 1.1) it reports 28.4% to 52.2% fewer output tokens and 34.4% to 51.6% lower average cost per task, with pass rates "within the confidence intervals of the base model". Its benchmark runs used a 512k-token limit.Source: particle.ai · checked 2026-11-17
  • Particle AI published the Particle-1 launch post on October 28, 2026, and says the model "is available today through the Particle API". The post gives no licence or open-weights information.Source: particle.ai · checked 2026-11-17
  • Particle’s model catalog gives particle-1 a 1,048,576-token context window (prompt plus completion), a default output cap of 32,000 tokens and an output ceiling of 131,072 tokens. It takes text and images in and returns text. Tool calling (OpenAI tools and tool_choice) is listed as "Qualified"; structured outputs are "Not advertised" (json_object is accepted, json_schema is reduced to a subset with no schema guarantee).Source: docs.particle.ai · checked 2026-11-17
  • Particle’s reasoning guide says all its models think by default: leaving reasoning_effort out means "high". For Particle-1, "none" turns thinking off, "minimal" and "low" mean low effort, "medium", "high" and "xhigh" mean high effort, and "max" and "ultra" mean max effort. Reasoning tokens count as output tokens and need room in the output limit.Source: docs.particle.ai · checked 2026-11-17
  • Particle’s Usage & billing page lists particle-1 at $0.25 per million input tokens, $0.03 per million cache-hit input tokens and $1.00 per million output tokens as the peak (standard) price, 04:00 to 20:30 New York time. From 20:30 to 04:00 an off-peak price of 50% applies automatically: $0.125 / $0.015 / $0.50. The Flex service tier takes a further 50% off, with lower priority when the service is busy. The page says account-specific prices may apply.Source: docs.particle.ai · checked 2026-11-17
  • The launch post’s benchmark footnote lists different prices for Particle-1: $0.30 uncached input, $0.006 cached input and $1.20 output per million tokens (the same as it lists for DeepSeek V4.1 Flash). These do not match the $0.25 / $0.03 / $1.00 on Particle’s billing and model pages, so this page shows the billing-page figures and flags the gap; we could not tell which one is current beyond the docs calling theirs "current".Source: particle.ai · checked 2026-11-17
  • Particle’s docs describe an OpenAI-compatible API at https://api.particle.ai/v1 (or https://us.api.particle.ai/v1 for inference on US GPUs only); keys come from platform.particle.ai. Send the model name particle-1; responses return it as particle-ai/particle-1. The same API also serves DeepSeek V4.1 Flash and GLM 5.3 Flash, and the docs recommend particle-1 as the starting point for new integrations.Source: docs.particle.ai · checked 2026-11-17
  • Particle’s data-retention page says it operates zero data retention for inference traffic: prompts and completions are "processed in memory and never persisted". It does store usage records (token counts, model, cost, no content), uploaded files until you delete them, and short-lived prefix caches. This describes Particle’s own API; chats on this page go through OpenRouter, whose handling we did not check here.Source: docs.particle.ai · checked 2026-11-17
  • OpenRouter’s model list shows particle-ai/particle-1 ("Particle AI: Particle-1", created October 28, 2026) at $0.06 input and $0.50 output per million tokens with $0.003 cache read, a 1,048,576-token context, a 131,072-token maximum output and text and image input; its description calls it a "token-efficient agentic coding model" post-trained on DeepSeek V4.1 Flash. The list also has a separate variant, particle-ai/particle-1:free, at no token charge; this page does not use it, and we did not check its limits. OpenRouter’s $0.06 / $0.50 is below Particle’s own peak price of $0.25 / $1.00 (checked November 17, 2026).Source: openrouter.ai · checked 2026-11-17 · second-hand

Nearest available

Models you can use now instead

$0.30 / $1.2 (per million tokens)

  • DeepSeek released V4.1-Flash on September 10, 2026.Source: api-docs.deepseek.com · checked 2026-11-11
  • Only the Flash version of V4.1 exists: DeepSeek has published no V4.1 Pro, and the Pro model on its price page is still deepseek-v4-pro (V4-Pro-0813).Source: api-docs.deepseek.com · checked 2026-11-11
  • DeepSeek’s pricing page says the API model deepseek-flash is now served by the V4.1-Flash-1017 snapshot.Source: api-docs.deepseek.com · checked 2026-11-11

FAQ

Questions

What is Particle-1?

Particle-1 is a model from Particle AI, launched on October 28, 2026. Particle says it is "post-trained on DeepSeek V4.1 Flash" with reinforcement learning so it solves tasks with fewer tokens, and that this first release focuses on software engineering. Its API docs recommend it for chat, agentic coding and tool calling. It takes text and image input and returns text.

How much does Particle-1 cost?

Particle’s own API lists $0.25 per million input tokens, $0.03 per million cache-hit input tokens and $1.00 per million output tokens at peak (04:00 to 20:30 New York time), and half of that off-peak (20:30 to 04:00); the Flex tier takes another 50% off. The launch post’s footnote shows $0.30 / $0.006 / $1.20, which does not match. OpenRouter lists particle-ai/particle-1 at $0.06 input and $0.50 output (checked November 17, 2026), and also a separate variant with the id ending :free that we did not check. On this site a typical reply to this model costs about 1 credit (a credit is about $0.01 in the $5 pack), billed by its real token cost, at least 1 credit.

How do I use Particle-1?

Through the Particle API (OpenAI-compatible, base URL https://api.particle.ai/v1, model name particle-1, key from platform.particle.ai), through OpenRouter as particle-ai/particle-1, or in the chat box on this page, on credits. Particle’s docs show how to point standard OpenAI SDKs at its base URL.

What is the context window of Particle-1?

Particle’s model catalog gives 1,048,576 tokens, prompt and completion combined. The default output cap is 32,000 tokens when you send no limit, and an explicit limit above 131,072 tokens is cut to 131,072. OpenRouter shows the same context and a 131,072-token maximum output.

How is Particle-1 different from DeepSeek V4.1 Flash?

Particle says Particle-1 is DeepSeek V4.1 Flash with a small trained adapter (rank-64 LoRA, about 1.25% of the parameters) that makes it reason with fewer tokens. In its own tests (not tested by us) it used 28.4% to 52.2% fewer output tokens on four coding and data benchmarks, with pass rates within the base model’s confidence intervals. Particle’s billing page lists the same $0.25 / $1.00 for both models. See DeepSeek V4.1 Flash.

Does Particle-1 think before it answers?

Yes, by default. Particle’s docs say Particle-1 reasons at high effort unless you set reasoning_effort, and that "none" turns thinking off. Reasoning tokens count as output tokens, so a thinking reply costs more than a short one. The chat on this page sends reasoning effort "none".

Is Particle-1 open source?

We could not confirm it. Particle’s launch post and docs give no licence or open-weights information, and describe Particle-1 as served through the Particle API. We did not find a weights page, so do not assume you can download it.

Is this the official Particle AI website?

No. This is an unofficial page, not affiliated with or endorsed by Particle AI. Facts link to Particle’s own pages where possible.