CA3 lets your open model keep learning after deployment, by your rules. What it learns is kept alongside the original, which never changes.
Every use case, grouped by type of model: from 22 million to 122 billion parameters, by size and by who makes it, all of them open-weight. The hollow dot ran without CA3.
Language models · LLMs
AI cybersecurity
gpt-oss-120b (OpenAI)
Biome Makers · client memory
Qwen3.5 122B-A10B (Alibaba)
Biome Makers · farm reports
Qwen3.8 27B (Alibaba), without CA3
Self-healing
Qwen3.5 0.8B (Alibaba)
Language
Qwen2.5 0.5B (Alibaba) · Qwen2.5 1.5B (Alibaba) · SmolLM2 1.7B (Hugging Face) · Qwen2.5 7B (Alibaba) · Llama 3.1 8B (Meta) · Qwen3 8B (Alibaba) · Qwen3 14B (Alibaba) · Nemotron 3 Nano 30B (NVIDIA)
AI judges
Strands Decider 2B (AWS) · Qwen3.5 9B (Alibaba)
Model edits
Gemma 3 270M (Google) · Qwen3.5 0.8B (Alibaba)
Flip-flop
Qwen3.5 0.8B (Alibaba)
Home robot
Qwen3 1.7B (Alibaba)
Personal app
LittleLamb 0.3B (Multiverse) · Qwen3 4B (Alibaba)
Home library
Qwen3 4B (Alibaba)
Speech and audio models
Speech
Granite Speech 4.1 2B (IBM)
Sound
MMAudio L 44k (1.03B, Sony AI · UIUC)
Vision and image models
Vision
DINOv2-small (22.1M, Meta)
Image generation
Krea 2 (12B, Krea)
Short film
Krea 2 (12B, Krea)
Earth-observation foundation model
Satellite
Prithvi-100M (NASA · IBM)
Robot foundation model · vision-language-action
Robot arm
MolmoAct2 (5B, Ai2)
Neural decoder · brain implant
Brain implant
SPINT decoder, size unknown (not given in its paper)
Recommender
Recommendation
Matrix factorisation, our own small model, under 1 million parameters
Model size, parameters, log scale
Sizes as published by each model's maker; point at a dot for its name and size. The hollow dot is the first phase of the Biome Makers pilot: the open model wrote the farm reports on its own, without CA3. The brain-implant decoder (SPINT) has no size in its paper, and the recommender is our own matrix factorisation, under 1 million parameters, below the axis: their rows carry no dot.
open-weight, from 22 million to 122 billion parameters. On your servers or ours.
See the models → Safe0of 24planted falsehoods kept. A RAG over the same news repeated 8.
See the case → Fast8×faster answers than the same model without CA3: 0.51 s against 4.03 s, in a personal app.
See the case → Accurate+43pointson six new European regions in six days, for NASA and IBM's crop model.
See the case →Retrain it, and it can break what already worked. Add RAG, and it pays to reread the same context on every question. Either way, nobody sees what changed.
Learns every day.
Keeps what it knew.
Blocks bad data.
Forgets on request.
You decide.
From agents that pass each other notes, AI judges that learn and a 122-billion-parameter model on a company's client data to a brain-implant decoder: language, robots, satellites, images, speech, vision, sound and recommendation.

The model heals itself; CA3 makes sure nothing else breaks. An AI edit made a small model six times better at math and quietly broke a skill. With CA3 as its controller, the model found the damage and healed it: 76.3% → 98.5%, with the math gain untouched.
Alibaba Qwen3.5-0.8B · edit by GPT-6 Astra, MIT CSAILRead the article →
Agents passing notes, like the ones that attacked Hugging Face. A rigged note fools the next agent 4 times in 10; with today's defenses, 3 or 4; with CA3 checking every note, 1.
OpenAI gpt-oss-120b · open weights · one GPUSee the case →
Real data of two clients. CA3 keeps one memory per client and region, learns new sheets on top and removes one on request. Account questions go to a program that computes every figure: none wrong in 130 answers, against 72 for document search.
Alibaba Qwen3.5-122B-A10B · open weights · two GPUsSee the case →
More right than RAG, at a seventh of the compute. Not one planted falsehood kept.
Alibaba Qwen · Meta Llama · SmolLM2 · NVIDIA Nemotron · 0.5B to 30BSee the case →
Every new fact a model learns is checked. A small judge learns from the big one to settle two thirds of the checks, and loses none of the 42 planted falsehoods the big one stopped.
Qwen3.5-9B and Strands Decider 2B · open weights · one GPUSee the case →
Seven new days learned; the earlier days kept at 0.939, fine-tuning 0.712.
SPINT decoder · FALCON public recordingsSee the case →All ten everyday tasks kept; ordinary fine-tuning lost them all.
Ai2 MolmoAct2 · simulatedSee the case →
+43.2 points in six regions; the US held at 60.
NASA and IBM Prithvi-100MSee the case →
One new character a day; none forgotten, none mixed up.
Krea 2 · research demonstrationSee the case →
7 weight edits written by an AI: 4 held back, each with what it breaks. 3 of those 4 had passed the usual benchmarks.
Alibaba Qwen3.5 · Google Gemma 3 · edits written by an AISee the case →
The original and an AI edit in one model, flipped in an instant; back to the original, bit for bit.
Alibaba Qwen3.5-0.8B · CA3 Flip-FlopSee the case →
Names 8% → 97%, and its English kept.
IBM Granite Speech 4.1See the case →
41% → 87% on it, the old camera kept, a planted trigger refused.
Meta DINOv2See the case →
39.8 MB per client, on one shared model.
Sony MMAudio · research demonstrationSee the case →
Every fake account's ratings held back.
Matrix factorisation · MovieLens-1MSee the case →Four to eight weeks on your own data, under NDA, with the same report.
Contact →Everyday things that learn a family: a robot vacuum, a coach app, a bookcase and a film with the toys of the house.

Twelve toys from the family's shelves, learned from phone photos by CA3 on Krea 2, come alive for one night. Premiere on 19 December at 19:00 (Spain), live on YouTube; its song, «Juguetito», is already on Spotify.
Krea 2 with CA3 · Wan 2.1 VACE · Seedance 2.5See the film →
544 real books learned from one photo per shelf: where each book is, 40 of 42, and new books found by their plot, 10 of 12. The same model untouched: 0.
Alibaba Qwen3-4B · a real bookcase · simulated weeksSee the case →Every lesson learned by the next morning; five bad ones refused, two forgotten on request.
Alibaba Qwen3-1.7B · Roborock S7 · simulated nightsSee the case →
Better advice than the same model, eight times faster.
Alibaba Qwen3-4B · Multiverse LittleLambSee the case →«El gran vuelo de Twiti», on 19 December 2026 at 19:00 (Spain), on the CA3 Labs channel.
Open the channel →The routine it settles on its own. The doubtful cases come to you, with the evidence side by side, and every decision stays on record.

229 of 3,136 checks, on three language models · every day a version: 1 of 29 undone on its own · See the case
From the language test: Alibaba Qwen3 14B, round 20d; moving to a new base model, Qwen2.5 7B to Qwen3 8B Base, round 13 · See the case
Four to eight weeks on your own data, under NDA. You get the same report on your data: accuracy and cost against your RAG, what it learned, what it refused and what it forgot.