XCoffee
Back to blog

Particle-1 API: Pricing, Benchmarks vs DeepSeek V4.1 Flash

Oct 30, 2026XCoffee Team

Particle-1 API guide: official price ($0.25/$1.00 per 1M), OpenRouter's lower listing, 1M context, vendor benchmarks vs DeepSeek V4.1 Flash, and how to call it.

Particle-1 API: Pricing, Benchmarks vs DeepSeek V4.1 Flash

Particle-1 API: Pricing, Benchmarks and How to Call It

Short answer: The Particle-1 API gives you Particle AI's agentic coding model, which is post-trained on DeepSeek V4.1 Flash to finish tasks in fewer tokens. On Particle's own API it costs $0.25 per 1M input tokens and $1.00 per 1M output tokens at peak, and half that off-peak. OpenRouter currently lists it at $0.05 / $0.50 (checked Oct 31, 2026). The context window is 1,048,576 tokens. You can call it through Particle's OpenAI-compatible API or through OpenRouter.

What is Particle-1?

Particle AI announced Particle-1 on October 28, 2026. It took DeepSeek V4.1 Flash and post-trained it with reinforcement learning for software work: terminal workflows, codebase changes and data engineering. The goal is shorter reasoning and fewer tool calls at about the same accuracy as the base model.

Spec Particle-1 (model catalog)
Developer Particle AI
Base model DeepSeek V4.1 Flash
Model ID particle-ai/particle-1 (alias particle-1 on Particle's API)
Context window 1,048,576 tokens (prompt plus completion)
Max output 131,072 tokens. The default cap is 32,000 if you don't set one
Input / output Text and images in, text out
Reasoning On by default. reasoning_effort defaults to high
Tool calling Yes (tools / tool_choice)
Structured output json_object is accepted. Particle calls JSON-schema conformance best-effort, not guaranteed
Weights No public weights seen as of Oct 31, 2026. Offered as an API

Particle-1 API pricing

All prices are USD per 1M tokens and were checked on Oct 31, 2026. Check the linked pages before you budget.

Where you buy it Input Cached input Output Notes
Particle API, peak (usage & billing) $0.25 $0.03 $1.00 04:00–20:30 New York time
Particle API, off-peak $0.125 $0.015 $0.50 20:30–04:00 New York time. Applied automatically
Particle API, Flex tier 50% off that period's rate Only if your account qualifies. Confirm with the X-Particle-Service-Tier header
OpenRouter (model page) $0.05 $0.003 $0.50 OpenRouter's listed price

Why two prices? The $0.25/$1.00 rate is Particle's own default, as published in its docs. The $0.05/$0.50 rate is what OpenRouter charges. According to OpenRouter's endpoint data, the only provider behind that listing is Particle AI itself, and no discount is flagged. Neither side says whether the lower OpenRouter price is temporary, so expect it could change.

On Particle's API, reasoning tokens count as output. Hosted web search costs $2.50 per 1,000 searches. Billing uses prepaid credits, and a request on an empty balance returns 402 insufficient_credits.

Particle-1 vs DeepSeek V4.1 Flash

  • Price on Particle's API: The two models have the same price, $0.25 input and $1.00 output.
  • Vendor-reported efficiency: Particle tested on Terminal-Bench 2.1, ReactBench, ADE Bench and DeepSWE 1.1. It reports 28.4%–52.2% fewer output tokens and 34.4%–51.6% lower average cost per task, with pass@1 inside the base model's confidence intervals. Its summary chart shows a 72.57% task pass rate for Particle-1 and 71.15% for DeepSeek V4.1 Flash (source). These are Particle's own numbers. No independent reproduction has been published.
  • A small third-party test: In a three-task coding test by Cocode, Particle-1 finished faster and used fewer output tokens than DeepSeek V4.1 Flash. It is a small sample.
  • When to pick it: It is worth trying for long agent loops where token count drives your bill. For other work, test both models on your own tasks.

How to call the Particle-1 API

Particle's official API (OpenAI-compatible)

curl https://api.particle.ai/v1/chat/completions \
  -H "Authorization: Bearer $PARTICLE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"particle-1","messages":[{"role":"user","content":"Hello"}]}'

Sign in at platform.particle.ai, add credits, and create a key. Because reasoning is on by default, leave room for it in max_completion_tokens.

OpenRouter

curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"particle-ai/particle-1","messages":[{"role":"user","content":"Hello"}]}'

Particle-1 in Claude Code and Codex (via OpenRouter, untested)

Particle's docs list only OpenAI-style endpoints. The settings below follow OpenRouter's own docs. We have not tested them with Particle-1.

Claude Code. This follows OpenRouter's Claude Code integration guide. OpenRouter notes that Claude Code is only guaranteed to work with Anthropic's own models. Using a non-Anthropic model such as Particle-1 is at your own risk.

export ANTHROPIC_BASE_URL="https://openrouter.ai/api"
export ANTHROPIC_AUTH_TOKEN="$OPENROUTER_API_KEY"
export ANTHROPIC_API_KEY=""            # must be explicitly empty
export ANTHROPIC_DEFAULT_SONNET_MODEL="particle-ai/particle-1"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="particle-ai/particle-1"

Codex. This follows OpenRouter's Codex CLI guide (~/.codex/config.toml):

model_provider = "openrouter"
model = "particle-ai/particle-1"

[model_providers.openrouter]
name = "openrouter"
base_url = "https://openrouter.ai/api/v1"

[model_providers.openrouter.auth]
command = "sh"
args = ["-c", "echo $OPENROUTER_API_KEY"]

Want flat-rate coding instead? XCoffee

XCoffee plan details updated November 12, 2026.

XCoffee does not offer Particle-1, and you cannot call Particle-1 through XCoffee. For coding models in general, XCoffee offers 30-day passes with credit allowances:

  • Prices, daily allowances and rolling hourly caps vary by plan. Check the current plans and your account’s limits.
  • Choose a task group: xcoffee-coding, xcoffee-fast or xcoffee. The live group list shows current multipliers and some member models.

To set it up in Claude Code or Codex, follow our Claude Code cheap API guide. Web search and Codex multi-agent are not available through XCoffee.

Particle-1 API FAQ

How much does the Particle-1 API cost?

On Particle's own API, peak rates are $0.25 per 1M input tokens, $0.03 for cached input and $1.00 for output. Off-peak is half that. OpenRouter lists $0.05 input and $0.50 output (as of Oct 31, 2026).

Why is Particle-1 cheaper on OpenRouter?

OpenRouter lists its own price. The only provider behind it is Particle AI. Neither party has said whether the lower price is temporary.

What is Particle-1 based on?

It is based on DeepSeek V4.1 Flash, post-trained by Particle AI for token-efficient agentic coding.

What is Particle-1's context window?

1,048,576 tokens, with up to 131,072 output tokens.

Is Particle-1 open source?

As of Oct 31, 2026 we found no public weights. Particle offers it as an API.

Can I use Particle-1 in Claude Code?

OpenRouter documents an Anthropic-compatible endpoint for Claude Code. It also says the setup is only guaranteed with Anthropic models, and we have not tested it with Particle-1.

Can I use Particle-1 through XCoffee?

No. XCoffee does not serve Particle-1.