Buy
Market
🔥
Prediction Market

Lightchain AI outlines roadmap for agents and per-call payments

Lightchain AI will introduce agents and pay-per-call payments over three months starting November 2026.

19/11/2026 15:1421 min read

Open-weight models still cannot match the best proprietary systems, but they have become usable for real applications and are much nearer the frontier than before. Lightchain AI’s next three months focus on what continues to restrain open AI: agents, payment mechanisms and a publicly verifiable network.

Open-weight models remain inferior to the top closed systems, but the gap has narrowed considerably. In 2024, Epoch AI estimated that open models were roughly one year behind the cutting edge. The research group’s latest reading, released in May 2026, shows that since January 2026 the best open-weight models have lagged the strongest closed ones by an average of four months. Models such as OpenAI’s gpt-oss, Alibaba’s Qwen3 and Meta’s Llama are now solid enough for building production applications, and they can be freely downloaded, inspected and executed.

What proprietary AI still controls is everything that surrounds the model itself. A single entity runs the servers, handles billing and develops the agents and tools that end users interact with every day. Even an open-weight model is only partly open when one company serves it, because that firm can modify it, alter its pricing or deactivate it.

Lightchain AI is constructing the other facet as an open network. From November 2026 through January 2027, its plan introduces capabilities typical of closed platforms—AI agents, pay-per-call billing and developer tools—on a blockchain where independent operators host the models and every request leaves a traceable record.

An open network for open models

Lightchain AI is a layer-1 blockchain tailored for artificial intelligence. Independent operators, termed workers, run open models on their own GPUs. The chain selects the worker for each request, holds the payment in escrow and writes an onchain record for every job, including its job ID, transaction details and the worker who fulfilled it.

The network currently supports 14 open models, rangeing from small Llama and Gemma variants up to gpt-oss 120B and DeepSeek R1, each at a fixed per-job fee. Users access them through Lightchain Chat, the Lightchain Desktop beta and Lightchain Studio, an open-source code editor that incorporates a coding agent where every prompt is a chain job.

Next: agents that anyone can call

The major AI companies are competing to develop agents—AI services that handle tasks and cooperate with other software. Lightchain’s three-month development timetable brings agents onto its own network, built and tested on the test network first.

  • November: the first agent. Agent contracts go live on the test network, and the first agent, atlas.lcai, is created in a single transaction. Each agent receives its own onchain name, enabling apps and other AI systems to find it, call it and pay it. A public gateway lets users look up agents by name.
  • December: agents that can be called and paid. Apps pay an agent in LCAI for every call, via the network’s x402-inspired pay-per-call system, and each payment is linked to one specific request. One agent can delegate a task to another, and anyone can confirm that a reply came from the job it purports to come from. On December 12, a public demo calls an agent from Claude Code, Anthropic’s coding tool, and pays it from a script.
  • January: the mainnet rehearsal. An independent audit of the agent contracts is completed and its findings published. An official worker build ensures all operators use the same model files and settings, and the full mainnet launch of the agents is rehearsed on the test network.

AI work that can be verified

When using a closed AI provider, users must trust that the work was executed as claimed. On Lightchain, every AI job produces an onchain record of when it ran and which worker handled it, the network repeats jobs to verify responses, and any user can challenge an answer within 24 hours. The record shows that the work occurred, not what was requested: prompts are encrypted before they reach the chain. The plan introduces additional verifications:

  • Every testnet job checked. Starting in November, the disputer—the network’s verification service—re-runs every job on the test network, whereas on mainnet it repeats a subset. By January it is to cover each model the network supports.
  • A network that keeps running. Following a test network outage on November 8, the plan secures public endpoints and gives every service a backup connection. In December the test network spreads its validators across seven or more servers at three or more providers, so losing two servers no longer halts the chain.
  • Operator rewards anyone can verify. A grants program for operators who run the network’s GPUs and nodes is being built in public. Uptime is logged hourly, grant tiers are based on the jobs a worker actually served, and every grants round ends in a payout root—a single fingerprint of the entire payout list that anyone can recalculate.
  • Progress you can verify. Every deliverable is linked to a transaction hash or address on the public testnet explorer, and the team publishes a written update each week.

The dates in the plan are targets, and each step reaches mainnet only after its test network rehearsal.

The next gap is the platform

Open models are not yet equal to closed AI, but they are close enough to base applications on them. What remains is the platform: who serves the model, who receives payment and who can audit the work. Lightchain’s next three months are its strategy to eliminate that gap openly, using agents that anyone can call, per-call payments and a public ledger that enables anyone to verify each request.

About Lightchain AI

Lightchain AI is a layer-1 blockchain built for artificial intelligence. Independent operators stake LCAI and run open models such as gpt-oss, Qwen3 and Llama on their own GPUs. The chain assigns each request, holds the fee in escrow and preserves a record of every request that anyone can inspect. Live products include Lightchain Chat, Lightchain Desktop (beta), Lightchain Studio, the Lightchain Bridge and the Lightscan explorer. The LCAI DUNA, a Wyoming association, serves as the project’s legal entity.

Further details are available at: Website | X | Discord

Share to

Disclaimer: this article comes from third-party media and is provided for reference only. It does not constitute investment advice. Crypto and other financial products carry significant price volatility risk, so please make your own decisions carefully.

Related articles