Which LLM should you use? Token Monster automatically combines multiple models and tools for you

Token Monster, a brand new AI chatbot platform, has launched its alpha preview, aiming to alter how customers work together with massive language fashions (LLMs).

Developed by Matt Shumer, co-founder and CEO of OthersideAI and its hit AI writing assistant Hyperwrite AI, Token Monster’s key promoting level is its means to route person prompts to the most effective accessible LLMs for the duty at hand, delivering enhanced outputs by leveraging the strengths of a number of fashions.

There are seven main LLMs presently accessible by Token Monster. As soon as a person sorts one thing into the immediate entry field, Token Monster makes use of pre-prompts developed by iteration by Shumer himself to mechanically analyze the person’s enter, determine which mixture of a number of accessible fashions and linked instruments are greatest suited to reply it, after which present a mixed response leveraging the strengths of stated fashions. The accessible LLMs embody:

• Anthropic Claude 3.5 Sonnet
• Anthropic Claude 3.5 Opus
• OpenAI GPT-4.1
• OpenAI GPT-4o
• Perplexity AI PPLX (for analysis)
• OpenAI o3 (for reasoning)
• Google Gemini 2.5 Professional

In contrast to different chatbot platforms, Token Monster mechanically identifies which LLM is greatest for particular duties — in addition to which LLM-connected instruments could be useful reminiscent of internet search or coding environments — and orchestrates a multi-model workflow.

“We’re simply constructing the connectors to all the pieces after which a system that decides what to make use of when,” stated Shumer.

For example, it would use Claude for creativity, o3 for reasoning, and PPLX for analysis, amongst others. This strategy eliminates the necessity for customers to manually select the correct mannequin for every immediate, simplifying the method for anybody who needs high-quality, tailor-made outcomes.

Function highlights

The alpha preview, which is presently free to enroll in at tokenmonster.ai, permits customers to add a spread of file sorts, together with Excel, PowerPoint, and Docs.

It additionally contains options reminiscent of webpage extraction, persistent dialog periods, and a “FAST mode” that auto-routes to the most effective mannequin with out person enter.

On the coronary heart of Token Monster is OpenRouter, a third-party service that acts as a gateway to a number of LLMs, and into which Shumer has invested a small sum, by his admission.

This structure lets Token Monster faucet into a spread of fashions from totally different suppliers with out having to construct separate integrations for every one.

Pricing and availability

As of proper now, Token Monster doesn’t cost a flat month-to-month payment.

As an alternative, customers solely pay for the tokens they devour by OpenRouter, making it versatile for various ranges of utilization.

In line with Shumer, this mannequin was impressed by Cline, a instrument that allows high-spending customers to entry limitless AI energy, permitting them to attain higher outputs by merely utilizing extra compute assets.

Multi-step workflows produce richer LLM responses

Token Monster’s AI workflows lengthen past easy immediate routing.

In a single instance, the chatbot would possibly begin with a analysis part utilizing internet search APIs, go that knowledge to o3 for figuring out data gaps, then create an overview with Gemini 2.5 Professional, draft textual content with Claude Opus, and refine it with Claude 3.5 Sonnet.

This multi-step orchestration is designed to supply richer, extra full solutions than a single LLM would possibly be capable of generate alone.

The platform additionally contains the power to save lots of periods, with knowledge securely saved utilizing the open supply on-line database service Supabase. This ensures that customers can return to ongoing initiatives with out dropping their work, whereas nonetheless giving them management over what knowledge is saved and what’s ephemeral.

A non-traditional CEO

In a notable experiment, Token Monster’s management has been handed over to Anthropic’s Claude mannequin.

Shumer introduced that he’s dedicated to following each determination made by “CEO Claude,” calling it a take a look at to see whether or not an AI can handle a enterprise successfully.

“Both we’ve revolutionized administration eternally or made an enormous mistake,” he wrote on X.

Rising from the Reflection 70-B controversy

Token Monster’s launch comes lower than a 12 months after Shumer confronted controversy over his launch and supreme retraction of Reflection 70B, a fine-tuned model of Meta’s Llama 3.1 that was initially touted as probably the most extremely performant open supply mannequin on the planet, however which shortly grew to become topic to criticism and accusations of fraud after third-party researchers have been unable to breed its acknowledged efficiency on third-party benchmark assessments.

Shumer apologized and stated the problems have been born out of errors made attributable to pace. The episode underscored the challenges and dangers of fast AI growth and the significance of transparency in mannequin releases.

MCP integrations coming subsequent

Shumer stated his workforce on Token Monster can be exploring new capabilities, reminiscent of integrating with Mannequin Context Protocol (MCP) servers that enable web sites and corporations to have LLMs make use of their information, instruments, and merchandise to attain higher-order duties than simply textual content or picture era.

This could allow Token Monster to attach with a person’s inside knowledge and providers, opening prospects for it to deal with duties like managing buyer help tickets or interfacing with different enterprise methods.

Shumer emphasised that Token Monster remains to be very a lot in its early phases. Whereas it already helps a collection of highly effective options, the platform stays an alpha product and is predicted to see fast iterations and updates as extra customers present suggestions. “We’re going to maintain iterating and including issues,” he stated.

A promising experiment

For customers who wish to make the most of the mixed energy of a number of LLMs with out the effort of mannequin switching, Token Monster could possibly be an interesting selection. It’s designed to work for individuals who don’t wish to spend hours tweaking prompts or testing totally different fashions themselves, as a substitute letting the system’s automated routing and multi-step workflows deal with the complexity.

As Token Monster’s capabilities develop, it is going to be attention-grabbing to see how customers and companies undertake it — and the way its experiment with AI-led administration pans out. For now, it’s a promising addition to the quickly increasing panorama of AI chatbots and digital assistants.