aime-Reasoner-2B™: a compact reasoning model built for the classrooms the cloud forgot

aime today released aime-Reasoner-2B, a compact language model designed for the moments when the cloud is not an option. Most AI progress assumes one thing silently: that there is a server somewhere, waiting to respond. For schools in rural India, for field researchers in low-connectivity zones, for students who study after the router goes off, that assumption breaks everything. aime-Reasoner-2B was built for those moments.

What it is

aime-Reasoner-2B is a 2-billion-parameter reasoning model shaped through aime's multi-teacher distillation programme. It was not trained only on answers. It was trained on how answers are formed. Different reasoning styles were brought together, aligned and compressed into a single model small enough to run on a laptop CPU, an old Android tablet, or a Raspberry Pi tucked into a classroom shelf. What emerged is not just smaller. It is more deliberate. The specific teacher families, training corpus and distillation methodology are aime's proprietary IP.

Think Cache — reasoning that doesn't start from zero

Every time a standard model is asked a question, it begins from scratch — re-processing the mechanics of how to think before it can think at all. aime-Reasoner-2B introduces a different approach: Think Cache. A Think Cache is a pre-built reasoning prefix — a snapshot of structured reasoning state — baked into the model's KV cache before inference begins. Instead of warming up the reasoning engine on every prompt, the model starts already in reasoning mode.

• Faster first-token response on constrained hardware.
• Consistent reasoning posture across all queries.
• No dependency on large context windows to prime behaviour.
• Stable performance whether the model is answering its first question or its fiftieth.

The Think Cache is not a shortcut. It is the result of distilling thousands of high-quality reasoning traces into a reusable starting state — so that every student who asks a question gets the same quality of structured thinking, regardless of the device they are on.

Think Book — lessons that live on the device

A Think Cache tells the model how to reason. A Think Book tells it what to reason about. Think Books are structured offline knowledge bundles — domain-specific reasoning packs that can be loaded into aime-Reasoner-2B without any internet connection. Each Think Book contains curated problem sets and worked examples, subject-specific reasoning traces in mathematics, logic and science, pre-structured question-answer flows for guided learning, and adaptive difficulty scaffolding for self-paced study.

Think Books are designed to be created once, distributed freely, and consumed entirely offline. A teacher in a low-connectivity school can author a Think Book using a simple local tool, copy it to a USB drive, and deploy it across an entire classroom of devices running aime-Reasoner-2B. The model reads the Think Book. The student asks the question. The reasoning happens on the device. No API call. No latency. No monthly subscription. No data leaving the room.

“Most AI progress quietly assumes there is a server somewhere, waiting to respond. aime-Reasoner-2B was built for the moments when that assumption breaks — the rural classroom, the field site, the student studying after the router goes off. The cloud is not an option there, and it doesn't need to be.”

A fully offline AI tutoring infrastructure

Together, Think Cache and Think Book enable something that has not existed before at this scale: a fully offline AI tutoring infrastructure. A student opens their device. The model is already loaded. The Think Book for today's lesson is already there. They ask a question about fractions, or logic, or percentage problems. The model works through the answer step by step — not because it is retrieving a cached response, but because it has internalised the structure of good reasoning and has the subject context it needs to apply it. The path to the answer is visible. The student can follow it. That is the point.

What it enables

• Offline tutoring — full reasoning support with zero connectivity.
• Think Cache deployment — pre-warmed reasoning state for consistent behaviour on any device.
• Think Book authoring — teachers create and distribute subject packs locally.
• Explainable answers — every response shows its reasoning chain, not just the conclusion.
• Low-resource hardware — runs on CPU-only devices including older laptops and single-board computers.
• Classroom scale — one USB drive can deploy the full system across an entire school.

“We didn't train this model only on answers. We trained it on how answers are formed. That is what makes a 2B model usable for teaching — not the size, but the discipline of the reasoning baked into it.”

A different kind of infrastructure

Most AI infrastructure is designed to scale up — more GPUs, more bandwidth, more compute. aime-Reasoner-2B is designed to scale out — to reach the places that larger systems cannot. It does not require a data centre. It does not require a stable connection. It does not require a procurement budget that most schools do not have. It requires a device, a Think Book, and a student with a question. That is enough.

“Every student who gets a step-by-step explanation instead of a blank screen is a student who learned something today. Every teacher who authors a Think Book is extending the reach of structured reasoning into a classroom that the cloud forgot.”

Who this is for

• Educators building offline lesson content for low-connectivity classrooms.
• EdTech developers deploying AI tutoring in emerging markets.
• Researchers studying reasoning behaviour in small models.
• Institutions with air-gapped or bandwidth-constrained environments.
• Students who learn best when the reasoning is shown, not just the answer.

Availability

aime-Reasoner-2B is available from today, May 5, 2026, alongside the Think Cache and Think Book architecture, for ministry, university and ecosystem partners deploying aime in offline and low-connectivity environments.

About aime

aime builds the operating system for educational intelligence — the foundational infrastructure layer that future education systems will run on. aime's stack combines structured curriculum knowledge, pedagogy-aware reasoning, compact education-tuned models, agentic orchestration and offline-capable deployment, and is designed for ministries of education, universities and national school systems.