Particle-1: Fewer Tokens for Software Engineering, Peak to Flex Pricing
Contains affiliate links; your price doesn't change · Disclosure · How we research
TL;DR
Particle-1 is a software engineering reasoning model post-trained on DeepSeek V4.1 Flash. Particle says it uses about 45% fewer tokens on average across tasks. Peak costs $0.25 per million input tokens, $0.03 cached and $1 output. Off-peak halves these rates. Flex takes another 50% off eligible token prices. Hosted search costs $2.50 per 1,000 searches and gets no Flex discount.
How this article was made: This article is research-based. Product claims come from Particle's launch post. Prices come from its Usage and billing documentation, checked on November 6, 2026 at 10:23 Taipei time. We have not called the API or rerun the benchmarks. Display name: CommX.
Particle-1 is a software engineering reasoning model post-trained on DeepSeek V4.1 Flash. Particle released it on October 28, 2026.
Particle says it uses about 45% fewer tokens on average across a wide variety of tasks. That is a vendor claim, not our test result.
Peak, Off-peak and Flex pricing
Prices are in USD per million tokens. These are default rates. Account-specific prices can differ.
| Tier | When it applies | Input | Cached input | Output |
|---|---|---|---|---|
| Peak | 04:00 to 20:30 America/New_York | $0.25 | $0.03 | $1.00 |
| Off-peak | 20:30 to 04:00 America/New_York | $0.125 | $0.015 | $0.50 |
| Flex | Effective service_tier is flex |
Another 50% off eligible token rates | Another 50% off | Another 50% off |
| Hosted web search | Per successful search call | $2.50 per 1,000 searches | Separate from token charges | No Flex discount |
The time window follows daylight saving time. The period is selected after routing and stays fixed for that request.
Flex has lower priority. Requesting it does not guarantee it. Check the X-Particle-Service-Tier response header before using the lower rate in an estimate.
Use the Usage and billing page for current prices. The launch post has older benchmark footnote prices. Do not use those for a current bill.
Check Particle-1 usage and billing
Where it fits
Particle-1 is a model API, not a host for your agent. Your application still needs somewhere to run if it must stay online. Our Zeabur shutdown migration guide covers that deployment side.
For more agent infrastructure notes, see OpenMail, which covers the separate problem of giving an agent its own email inbox.
Sources (checked November 6, 2026, Taipei)
- https://www.agentcomm.net/zh/particle-1-token-efficiency/
- https://particle.ai/blog/introducing-particle-1
- https://docs.particle.ai/usage-billing
- https://docs.particle.ai/service-tiers
FAQ
What is Particle-1?
Particle-1 is a software engineering reasoning API from Particle, released on October 28, 2026. It is post-trained on DeepSeek V4.1 Flash.
How much does Particle-1 cost?
Peak rates are $0.25 per million input tokens, $0.03 cached input and $1 output. Off-peak is $0.125, $0.015 and $0.50. An effective Flex tier halves eligible token rates again.
Does Flex reduce the search fee?
No. Hosted web search costs $2.50 per 1,000 searches by default. Flex discounts eligible model tokens, not the search tool fee.
Did CommX verify the 45% token saving?
No. About 45% fewer tokens on average is Particle's own claim across a wide variety of tasks. We did not rerun its benchmarks.
Key takeaways
- Particle-1 is a software engineering reasoning model post-trained on DeepSeek V4.1 Flash.
- Particle says it uses about 45% fewer tokens on average across tasks.
- Peak costs $0.25 per million input tokens, $0.03 cached and $1 output.
SourcesUsage and billing page · particle.ai/blog/introducing-particle-1 · docs.particle.ai/service-tiers
Last updated
Cite this pageParticle-1: Fewer Tokens for Software Engineering, Peak to Flex Pricing — https://www.agentcomm.net/articles/particle-1-token-efficiency/