PrismML

82 posts

PrismML banner
PrismML

PrismML

@PrismML

Centering AI research on efficiency. https://t.co/88MQHGCeFD

United States Katılım Mart 2025
24 Takip Edilen9.9K Takipçiler
Sabitlenmiş Tweet
PrismML
PrismML@PrismML·
Today, we are emerging from stealth and launching PrismML, an AI lab with Caltech origins that is centered on building the most concentrated form of intelligence. At PrismML, we believe that the next major leaps in AI will be driven by order-of-magnitude improvements in intelligence density, not just sheer parameter count. Our first proof point is the 1-bit Bonsai 8B, a 1-bit weight model that fits into 1.15 GBs of memory and delivers over 10x the intelligence density of its full-precision counterparts. It is 14x smaller, 8x faster, and 5x more energy efficient on edge hardware while remaining competitive with other models in its parameter-class. We are open-sourcing the model under Apache 2.0 license, along with Bonsai 4B and 1.7B models. When advanced models become small, fast, and efficient enough to run locally, the design space for AI changes immediately. We believe in a future of on-device agents, real-time robotics, offline intelligence and entirely new products that were previously impossible. We are excited to share our vision with you and keep working in the future to push the frontier of intelligence to the edge.
PrismML tweet media
English
177
592
4.2K
1.3M
PrismML retweetledi
Tushar Bansal
Tushar Bansal@tushar_bans·
We’re looking for people who love building from scratch! DM if interested.
Omead Pooladzandi@HessianFree

Hey! @PrismML is hiring! We're looking for LLM people who have trained models at scale - SFT/RL, data mixtures, evals, distillation, long context, distributed training, kernels, you name it! Especially interested in people who like owning the full stack from training dynamics -> shipped models. btw, we need a DevRel too. DM me.

English
11
7
116
12.7K
PrismML retweetledi
Evan Walters
Evan Walters@evaninwords·
My job is literally to do interesting and crazy experiments, if you are into that sort of thing DM Omead or I (or @pashakho, @SahinLale, @kmattar1981, @tushar_bans, @eraznafre, @NMonti25537) and come build with us at @PrismML!
Omead Pooladzandi@HessianFree

Hey! @PrismML is hiring! We're looking for LLM people who have trained models at scale - SFT/RL, data mixtures, evals, distillation, long context, distributed training, kernels, you name it! Especially interested in people who like owning the full stack from training dynamics -> shipped models. btw, we need a DevRel too. DM me.

English
3
2
39
4.5K
PrismML retweetledi
Sahin Lale
Sahin Lale@SahinLale·
We’re expanding our highly technical team at @PrismML — people who love pushing model quality end-to-end, from training dynamics to shipped models. If you’ve scaled LLM training, RL/SFT, evals, distillation, long context, kernels, or infra, we’d love to talk.
Omead Pooladzandi@HessianFree

Hey! @PrismML is hiring! We're looking for LLM people who have trained models at scale - SFT/RL, data mixtures, evals, distillation, long context, distributed training, kernels, you name it! Especially interested in people who like owning the full stack from training dynamics -> shipped models. btw, we need a DevRel too. DM me.

English
4
5
22
2.4K
Omead Pooladzandi
Omead Pooladzandi@HessianFree·
Hey! @PrismML is hiring! We're looking for LLM people who have trained models at scale - SFT/RL, data mixtures, evals, distillation, long context, distributed training, kernels, you name it! Especially interested in people who like owning the full stack from training dynamics -> shipped models. btw, we need a DevRel too. DM me.
English
16
19
382
40.6K
PrismML retweetledi
Pete Soderling
Pete Soderling@petesoder·
One of my favorite things about running @AICouncilConf for eleven years? The founders. There's a secret "track" that's not on the schedule — an invisible hallway of builders. And the next wave is showing up at SF 2026: 🧵 @EnoReyes of @FactoryAI @vikhyatk of @moondreamai @ds3638 of @honeyhiveai Emilie Schario of @kilocode @ianlivingstone of @KeycardLabs @neilmovva of @sailresearchco @CompleteSkeptic of @typesafeai @HessianFree of @PrismML @latkins of @arcee_ai Iona Hreninciuc of @runware petesoder.substack.com/publish/post/1…
English
0
4
16
2.1K
PrismML retweetledi
AI Council
AI Council@AICouncilConf·
Training gets the headlines. Inference gets the bill. As agents move from novelty to default workload, the hard problem isn't the model anymore. It's every millisecond and every watt between a prompt and the next token. A coding agent running for six hours straight is a very different customer than a chatbot, and the serving stack built for one isn't the stack the other needs. The Inference Systems track at AI Council 2026 is where the people rebuilding that stack share what they're actually doing. Curated by @BEBischof, Head of AI at @Theoryvc. Here's the lineup: → @yaroslavvb, Principal Researcher at @togethercompute: "What Comes After Deep Learning?" → Neil Movva, Co-Founder at Sail Research: "Great Infra for Background Agents" → @vikhyatk, CTO at M87 Labs (@moondreamai): "No Dropped Frames: Designing a VLM Around a Latency Budget" → @johnpdickerson, CEO at @MozillaAI: "Do the Boring Stuff to Make Open Source AI Win" → @CompleteSkeptic, Co-founder & CEO at @typesafeai (co-inventor of ChatGPT): "AI: Too Good to Be True, Too Bad to Be Useful" → Sriram Vishwanath, Professor and Founder at @GeorgiaTech: "Beyond Next-Token Prediction: Joint Embeddings, World Models, and Why Natural Language Isn't Enough" → Omead Pooladzandi, Co-Founder & Co-Head of Research at @PrismML: [talk title forthcoming] Thanks for curating a great track, Bryan! See you at AI Council 2026 in SF, May 12–14. 🎟️ aicouncil.com
AI Council tweet mediaAI Council tweet media
English
17
4
26
3.7K
PrismML retweetledi
Grover GPT
Grover GPT@GroverGPT·
Tiny local models like Bonsai are going to change things. For the last three years, the default way most people used AI was simple: frontier models lived in data centres, you reached them through an API, and anything local felt like a toy. That will probably stop being true in the near-ish future. Models are getting small enough to run on hardware people already own, cheap enough to live inside real power budgets, and good enough for a large share of everyday tasks. Bonsai is one recent example: Prism’s 1-bit releases are explicitly aimed at Apple hardware, including iPhone. If that continues, a lot changes. 1/ Cloud stops being the default for every task. You reach for it when the task actually justifies it. 2/ Privacy can become the default again, because more inference happens on your device instead of somewhere else’s server. 3/ The economics change. You stop paying frontier prices for every interaction forever, and start reserving expensive remote intelligence for the moments that need it. 4/ “Best model” splits in two: best in absolute terms, and best you can actually run on your phone. The energy part matters too, and maybe more than people think. The point is not just that local models may use less power for routine tasks. It is that they could change the energy topology of AI by pushing a large share of useful cognition onto hardware that already exists, including the NPU already sitting inside your device today. That changes the race. The question stops being only who has the smartest model in a lab. It becomes who can deliver the most useful intelligence under a real power budget. That rewards efficiency, distillation, open weights, clever hardware, and products that know when local is enough and when the cloud is worth the cost. Cloud does not disappear. It becomes a premium tier of cognition. And that opens the door to a very different kind of AI company from the ones winning today.
PrismML@PrismML

Today we’re announcing Ternary Bonsai: Top intelligence at 1.58 bits Using ternary weights {-1, 0, +1}, we built a family of models that are 9x smaller than their 16-bit counterparts while outperforming most models in their respective parameter classes on standard benchmarks. We’re open-sourcing the models under the Apache 2.0 license in three sizes: 8B (1.75 GB), 4B (0.86 GB), and 1.7B (0.37 GB).

English
4
1
29
4.6K
PrismML retweetledi
Pico AI Server and Pico AI Studio
Pico AI Server and Pico AI Studio@PicoGPT·
Pico Local AI Server 1.4.21 is now available on the Mac App Store. This release adds support for Ternary Bonsai, a lightning-fast model that outperforms many much larger models
PrismML@PrismML

Today we’re announcing Ternary Bonsai: Top intelligence at 1.58 bits Using ternary weights {-1, 0, +1}, we built a family of models that are 9x smaller than their 16-bit counterparts while outperforming most models in their respective parameter classes on standard benchmarks. We’re open-sourcing the models under the Apache 2.0 license in three sizes: 8B (1.75 GB), 4B (0.86 GB), and 1.7B (0.37 GB).

English
2
2
11
2.3K
PrismML retweetledi
Mustafa Ergisi
Mustafa Ergisi@mustafaergisi·
@PrismML Ran Ternary-Bonsai 8B on my iPhone through OnDevice LLM. Surprisingly fast.
English
2
6
22
3.8K
PrismML retweetledi
Jon Durbin
Jon Durbin@jon_durbin·
Ternary is actually surprisingly powerful. Validated by bitnet and now again here. In the new model training research/experimentation I've been working on, ternary weights (in some places) actually beats bf16 (by a not-insignificant amount), at least up to the 7b scale (and with every indication that this benefit scales up).
PrismML@PrismML

Today we’re announcing Ternary Bonsai: Top intelligence at 1.58 bits Using ternary weights {-1, 0, +1}, we built a family of models that are 9x smaller than their 16-bit counterparts while outperforming most models in their respective parameter classes on standard benchmarks. We’re open-sourcing the models under the Apache 2.0 license in three sizes: 8B (1.75 GB), 4B (0.86 GB), and 1.7B (0.37 GB).

English
4
8
82
8.8K
PrismML retweetledi
Michel aka Agent B
Michel aka Agent B@MichelIvan92347·
Interesting work here 👇
PrismML@PrismML

Today we’re announcing Ternary Bonsai: Top intelligence at 1.58 bits Using ternary weights {-1, 0, +1}, we built a family of models that are 9x smaller than their 16-bit counterparts while outperforming most models in their respective parameter classes on standard benchmarks. We’re open-sourcing the models under the Apache 2.0 license in three sizes: 8B (1.75 GB), 4B (0.86 GB), and 1.7B (0.37 GB).

English
0
4
7
2.5K
PrismML retweetledi
Kanu Gulati @Khosla Ventures
Kanu Gulati @Khosla Ventures@KanuGulati·
Ternary Bonsai 8B is within 5% of Qwen 3 8B at 9x lower memory! Congratulations @PrismML on yet another exciting release! cc @khoslaventures
PrismML@PrismML

Today we’re announcing Ternary Bonsai: Top intelligence at 1.58 bits Using ternary weights {-1, 0, +1}, we built a family of models that are 9x smaller than their 16-bit counterparts while outperforming most models in their respective parameter classes on standard benchmarks. We’re open-sourcing the models under the Apache 2.0 license in three sizes: 8B (1.75 GB), 4B (0.86 GB), and 1.7B (0.37 GB).

English
2
8
39
5.6K
PrismML retweetledi
0xSero
0xSero@0xSero·
One of the things I tried researching but found really hard. 1.58bpw is insane 10x smaller than original, I hope they push it to much larger models
PrismML@PrismML

Today we’re announcing Ternary Bonsai: Top intelligence at 1.58 bits Using ternary weights {-1, 0, +1}, we built a family of models that are 9x smaller than their 16-bit counterparts while outperforming most models in their respective parameter classes on standard benchmarks. We’re open-sourcing the models under the Apache 2.0 license in three sizes: 8B (1.75 GB), 4B (0.86 GB), and 1.7B (0.37 GB).

English
9
11
208
12.6K
PrismML retweetledi
Sahin Lale
Sahin Lale@SahinLale·
Check out how you can use Ternary Bonsai 8B 🌳 for tool calling in your everyday life—an impressive demo on an amazing platform by @AnythingLLM and @tcarambat!
English
1
8
20
2.2K
PrismML retweetledi
Robert Scoble
Robert Scoble@Scobleizer·
The models are getting smaller. Great for OpenClaws and Hermes. Gotta heat them up! Yesterday someone told me "phones are three to five years away." Oh, really?
PrismML@PrismML

Today we’re announcing Ternary Bonsai: Top intelligence at 1.58 bits Using ternary weights {-1, 0, +1}, we built a family of models that are 9x smaller than their 16-bit counterparts while outperforming most models in their respective parameter classes on standard benchmarks. We’re open-sourcing the models under the Apache 2.0 license in three sizes: 8B (1.75 GB), 4B (0.86 GB), and 1.7B (0.37 GB).

English
6
11
75
9K
PrismML retweetledi
Xenova
Xenova@xenovacom·
Ternary Bonsai: state-of-the-art intelligence at 1.58 bits. The models are so small they can even run locally in your browser on WebGPU! ⚡️ Here's the 8B version (just ~2GB in size) running at 60 tokens per second on my M4 Max. Try the demo out yourself! 👇
English
8
35
188
23.4K

Keşfet

@SailResearchCo @MoondreamAI @Mozilla @TypeSafeAI @TogetherCompute @BEBischof @TheoryVC @pashakho