Pinned Tweet
1/ Excited to announce our new company Particle
This is our first step in a series of models focused on developers.
With particle-1, we focused on token efficiency and managed to keep parity in performance while using ~45% fewer tokens.
Live now: particle.ai
“Once man has created an intelligence greater than his own, that intelligence should be able to create an intelligence greater than itself, as man whose intelligence is less than its was able to create its intelligence.”
One reason I am sanguine about the technological singularity is that I have seen it coming for over 50 years. In 1975, when I was 17, I wrote:
"Before man can most effectively aid his problem solving machinery, he must first understand it. At present, his understanding is very poor. However, when the breakthrough in understanding the brain occurs, improvement will come very fast and in an exponential increase. Once man has created an intelligence greater than his own, that intelligence should be able to create an intelligence greater than itself, as man whose intelligence is less than its was able to create its intelligence. Thus, the level of intelligence will shoot up with incredible rapidity, probably within the space of years. Within a few years of a breakthrough, which may occur any time now, man will either have been left behind by a superior intelligence of his own making, which he can no longer understand, or he will have become that intelligence and will no longer be like anything imaginable today."
Easily the coolest job for a hardware person.
We're doubling down on hardware and are looking for someone to manage our hardware shop ycombinator.com/companies/y-…
Excited to be partnering with Cloudflare and on their AI gateway - try our new token efficient model. 🚀
Particle-1 from @particle_api is available on AI Gateway — a model designed to be token-efficient, fast, and accurate. The hottest models from the hottest new labs, all on @Cloudflare AI Gateway.
developers.cloudflare.com/ai…
Independently validated in real-swe!
we benchmarked particle-1 on realswe.com
turns out you can achieve the same performance with 43% fewer tokens
congrats @Suhail
Suhail retweeted
Introducing Phonic-1, the most advanced speech-to-speech model for agentic tasks — from @phonic_co, the speech-to-speech model company.
We’ve had a breakout year and raised a $27.5M Series A led by @firstround with participation from @Lux_Capital, @felicis, and Definition.
I am very excited to see someone independently validate our claims on something we couldn’t train on and is real-world.
Same performance capability, with way less tokens.
Particle-1 is #2 for efficiency on the Parallel Search Capability Leaderboard.
It matches its base model, DeepSeek V4.1 Flash, on web research with Parallel Search (69.2 vs. 68.9) while using 42% fewer output tokens.
Suhail retweeted
We ran particle-1 through our Hacking Race today.
While it was trained for coding and data work, not security, it still got working exploits on 2 of our 4 targets, with zero false positives on the patched versions.
Strong first model from @Suhail and the particle.ai team
Particle's first model, particle-1, verified 4 of 11 vulnerable runs in our AI Hacking Race, and its four verified exploits cost about $1 combined.
It solved the Grafana plugin import and Jenkins command line parser tasks. The median verified run took about 4.7 minutes of active model time.
enclave.ai/hackingrace
Suhail retweeted
pretty impressive results on ReactBench!
1/ Excited to announce our new company Particle
This is our first step in a series of models focused on developers.
With particle-1, we focused on token efficiency and managed to keep parity in performance while using ~45% fewer tokens.
Live now: particle.ai
Very excited to announce partnering with @OpenRouter and launching our Inference service.
particle-1 is live! Please send us feedback. Will respond to all DMs if you find any issues.
Happy to fix things.
Fewer tokens, same work. Particle-1 from @particle_api uses 28–52% fewer output tokens than DeepSeek V4.1 Flash across four coding benchmarks, per Particle. Now on OpenRouter.
Boom 🚀
Excited to partner with Vercel and @rauchg
Particle-1 is now available on AI Gateway.
𝚙𝚊𝚛𝚝𝚒𝚌𝚕𝚎/𝚙𝚊𝚛𝚝𝚒𝚌𝚕𝚎-𝟷
Post-trained on DeepSeek V4.1 Flash, it completes engineering tasks with fewer tokens. Use it to debug, plan code changes, and break down multi-step problems. vercel.com/changelog/particl…
4/ Our mission is to make intelligence as accessible and abundant as electricity.
If you’re interested in low-level inference optimization, autonomous post-training, and the joining our rebel army to decrease the concentration of AI, please consider reaching out to us—we’re hiring!
Join us: particle.ai/about
I want to thank a bunch of people who helped educate/gave me feedback me on things:
- @levie hounding me every day and encouragement to chase this idea
- @rahulvohra and @_arjun for their eyes on the brand
- @jingli9111 @natolambert @jshobrook for letting me ask endless questions on model training to ramp up
- @gorkem @bavaria_kush for getting us compute when we needed it most
- @garrytan for helping me get that f*cking pebble out of my shoe last year
- @kyleanthony for the amazing brand work
- The @opencode guys for teaching us how high the bar is for inference and what great should look like
- all our investors for their instant belief in me
- ....and my team who took a leap of faith with their time + energy