AI Summary
Show More
Quickly grasp the article's content and gauge market sentiment in just 30 seconds!
Since ChatGPT's arrival on the scene, large language models (LLMs) have attracted the attention of artificial intelligence (AI) enthusiasts and developers. ChatGPT's impressive reasoning and text generation capabilities are underpinned by LLM technology, which has led to significant interest in these generative AI models. In the crypto space, LLM-focused projects like Bittensor (TAO) and Swarms (SWARMS) have also benefited from the widespread interest in this technology.
While LLMs are truly a revolutionary technology and, undoubtedly, the most popular algorithms in the AI field at the moment, another category of generative AI models — small language models (SLMs) — holds significant promise in many areas, thanks to domain-specific training and resource efficiency. While not as powerful and rich in generalist knowledge as LLMs, SLMs excel in situations where efficient access to specialized subject matter knowledge is critical.
Assisterr (ASRR) is a Solana (SOL)–based decentralized project that aims to leverage SLMs to help its users build efficient and knowledgeable AI agents, bots and solutions. Like LLM-based chatbots and models, Assisterr's SLM solutions utilize generative AI technology, but they feature much more scalable, resource-efficient algorithms that excel in situations in which specialized domain knowledge is essential. Furthermore, the use of the Mixture of Agents (MoA) concept allows Assisterr to draw on the advantages of multi-agent processing.
Anyone (including users without programming skills) can use Assisterr to create their own SLM-based agents and solutions. If SLMs attract even half the interest that LLMs currently enjoy, we’ll soon be hearing a lot more about the Assisterr crypto project, a true pioneer of the SLM branch of generative AI.
Key Takeaways:
Assisterr AI (ASRR) is a decentralized platform that lets you create AI agents and solutions based on SLMs — specialized AI models that feature resource efficiency and robust domain-specific knowledge.
Assisterr's native token, ASRR, is used for governance, payments to create and utilize the platform's products, model deployment, airdrop distributions and rewards to contributors.
You can buy ASRR on Bybit as a USDT Spot pair.
Assisterr (ASRR) is a decentralized AI project that lets you build SLM-based agents, without any programming skills. You can use a graphical interface to make selections and provide the necessary natural language prompts to design your agent and its properties. Assisterr is based on the Solana blockchain, so it benefits from the network's famed scalability and low costs.
The platform lets you build and deploy AI agents to track key metrics, identify patterns and extract valuable insights by analyzing large datasets in real time. For instance, an AI-powered trading agent can help you monitor market conditions, understand market dynamics and evaluate price dynamics. This enables you to refine your strategies and respond quickly to changing market sentiment. From forecasting trends to automating reports, AI agents empower users to stay ahead of rapid changes and optimize performance across industries.
The platform's easy-to-use and efficient AI agent framework benefits both personal and business users. For personal users, AI agents serve as powerful tools that help them automate routine tasks or perform sophisticated blockchain analysis. Individuals may also choose to make their custom agents public for monetization. Business users might wish to keep their agents private and utilize them for specific purposes, whereby specialized domain knowledge provides a key competitive advantage.
While most other AI projects utilize LLMs, Assisterr is one of the few web3 initiatives that leverages SLMs. While LLMs are famous for their breadth of expertise and excellent generalist knowledge, they’re also computationally intensive, and lack the necessary granularity regarding domain-specific topics. This is because LLMs are trained on massive amounts of data and use many parameters, making them resource-hungry and mostly generalist.
In contrast, SLMs use fewer parameters and significantly less data for training, reducing the overall computational requirements. At the same time, SLMs are typically trained on specific niche data, often making them better than LLMs in terms of deep specialist knowledge.
Just as importantly, many LLM-based projects are based on centralized servers, which are often controlled by a single corporate entity. As such, these models' exact training methods and data remain a huge unknown for end users. Assisterr's decentralized nature makes its entire model training process community-owned and more transparent.
Nik Havryliak and Dmytro Dimenko founded the Assisterr project in April 2023. In late May 2025, Assisterr was launched on the Solana mainnet.
Assisterr allows you to design your own SLM-based AI agents and deploy them to social media and blockchain sources. Each user can deploy up to three models.
There are three primary roles on the network:
Model creators are users who design, deploy and maintain SLMs.
Validators contribute their subject matter knowledge to approve and fine-tune the models.
Contributors are parties who contribute to the broader Assisterr ecosystem by participating in community discussions, providing feedback, sharing datasets or helping to promote the platform or specific models.
SLMs are the core technology that powers Assisterr AI agents. These models are resource-efficient, due to much lower training data and parameter loads than for LLMs. Their primary advantage is also domain-specific training.
Since SLMs are based on smaller datasets than LLMs, their performance is often augmented through multi-agentic collaborative efforts, whereby multiple models, agents or parallel learning networks are used to deliver a solution to user query. If a query requires multi-discipline knowledge, SLMs, with their domain specialization, may struggle in a single-model, non-collaborative setup. But the key limitations of SLMs can be resolved by using multiple models in a modular fashion to deliver the required quality of outputs.
Two architectural varieties are used to form the ensemble setup for SLMs — Mixtures of Experts (MoE) and Mixtures of Agents (MoA).
MoE is a model architecture that dynamically selects from a pool of specialized neural networks to solve each specific query. This selective activation allows for more efficient, content-aware problem solving. More complex MoEs can be layered hierarchically in order to tackle increasingly contextual complex problems.
MoA differs from MoE by utilizing the agents in a parallel manner. Each agent independently carries out specific tasks, based on its niche knowledge. The agents’ outputs are synthesized and aggregated by a separate model to produce the final result, and the aggregation of small models into more complex ones helps facilitate efficient problem-solving across domains.
All the models offered by the platform are based on decentralized, community-driven ownership. This starkly contrasts with the centralized model ownership and management usually found in the traditional LLM setups in the Web 2.0 world. The decentralization of model ownership helps solve many of the limitations of centralized systems. First, it provides more transparency around the data used to train a model. Secondly, it makes the model censorship-resistant. And thirdly, it helps align incentives for all parties involved, including Contributors, Validators and Creators.
Each SLM on Assisterr is managed through its own decentralized autonomous organization (DAO). This means there’s a dedicated treasury and governance mechanism for each model, a structure that helps maintain accountability and transparency for each SLM. It also makes model deployment, adjustment and maintenance more flexible and tied to the contextual requirements of each SLM.
Assisterr has an on-platform marketplace where model creators can list and tokenize their SLM-based solutions. An integral part of the marketplace is the Data Market, which facilitates the sharing and trade of datasets for model training. Data providers make curated datasets available through the market. The curated nature of the datasets ensures that model creators can rely on high-quality data to build their models, a critical consideration in an industry in which access to verified, quality data is becoming increasingly challenging.
Assisterr has partnered with several data providers in order to facilitate access to curated datasets for model creators. These data providers cover a range of industry and use case verticals. Some of the data partners on the platform include:
dFusion AI is a provider focused on quality data through community-driven validations. The company processes more than 50,000 knowledge submissions monthly, and has built a knowledge base of over 40GB, making it a crucial player for AI data in the web3 and blockchain niches.
FileMarket AI is a provider that specializes in collecting and annotating data with full user consent. At the core of FileMarket AI’s data solutions are ethical practices and privacy, which ensures that any data supplied by the company satisfies the most stringent regulatory standards.
Matchain is a platform specializing in decentralized identity (DID) solutions. It aggregates data from a variety of on-chain and off-chain sources and enriches the collected data using AI algorithms.
Assisterr’s native token, ASRR, was launched on the Solana blockchain in late May 2025. The ASRR token is a core asset within the Assisterr ecosystem, serving multiple purposes:
Transaction fee payments for the use of platform services. All operations on Assisterr — including listing models on the marketplace and paying for model computations — are paid in ASRR.
Model and treasury deployment fees. Launching individual SLMs and setting up their treasuries also involves payments in ASRR.
Governance. ASRR holders have the right to participate in platform-wide governance. Individual SLM project governance is carried out through management tokens (MTs), which are issued by creators specifically for each SLM project.
Rewards to project contributors. Contributors of various kinds, including developers, model builders and model validators, are rewarded in ASRR tokens.
Airdrop distributions. Platform contributors are rewarded with ASRR token distributions. Eligible contributors earn points for completing various tasks in order to qualify for the airdrops. For instance, the platform’s Stage 1 airdrop campaign provided ASRR rewards to SLM creators, validators, Assisterr NFT holders and other active contributors. Eligible parties accumulated sASRR points — the testnet reward asset of the platform — to unlock their ASRR allocations immediately upon the token generation event (TGE). The allocations were available immediately after the TGE, with no vesting schedules, conditional unlock schedules or other limitations.
ASRR’s maximum supply is 100 million, nearly all of which represents the token's current total supply.
The ASRR token is available on Bybit as a Spot pair with USDT. You can also enjoy ASRR staking rewards by participating in Bybit's event dedicated to the token. Under the terms of the campaign, you can stake between 160 and 3,000 ASRR via the 3-day Fixed Savings plan to earn 99% APR and grab a share of the 500,000 ASRR reward pool. This campaign is valid through Jul 3, 2025, 10AM UTC.
As of Jun 17, 2025, the ASRR token has been on the market for about two weeks. It’s currently trading at $0.25, a 48.7% drop from its all-time high of $0.5009 on Jun 3, 2025, and a 13.5% increase from its all-time low of $0.2264 on Jun 6, 2025.
ASRR’s long-term price outlook is quite bullish.DigitalCoinPrice expects the token to trade at $0.90 in 2027 and $1.31 in 2030, whilePricePrediction is forecasting average prices of $0.71 in 2027 and $2.11 in 2030.
In the sea of new AI-focused web3 projects, Assisterr AI stands out as a unique platform in the crypto industry because it leverages SLMs, rather than LLMs, to provide intelligent agents and solutions to its users. SLMs aren't an entirely new technology, but their use within the web3 and Web 2.0 domains has been limited.
This represents a perfect opportunity for the Assisterr platform to establish itself as one of the pioneers in leveraging these models. Whenever we hear about ChatGPT, we think of its creator, OpenAI. In the future, when SLMs are mentioned, we may be linking the technology to the Assisterr brand if its SLM-based models manage to gain traction in the AI market.
#LearnWithBybit