ArXiv Scholar vs ModelHub API

ArXiv Scholar and ModelHub API are both inference engines & infra tracked by AIDiveForge. Below is a side-by-side comparison of pricing, capabilities, platforms, and ownership — sourced from each tool's live website and verified before publishing.

ArXiv Scholar

ArXiv Scholar is an open-source RAG infrastructure that indexes roughly 5,600 curated AI engineering papers from arXiv and exposes them through a streaming API, so agents and developers can query verified literature instead of relying on a model's training memory. The retrieval pipeline runs a 1ms ML-based router that classifies each query as Direct, Decompose, or HyDE before spinning up hybrid dense-plus-sparse search and a cross-encoder re-ranker. Every answer ships with real arXiv paper IDs attached. The hard ceiling is the corpus: 5,600 papers covering RAG, LLMs, agents, training, and inference — nothing outside that domain, and nothing beyond what was ingested through the pipeline as of June 2026. The public endpoint is rate-limited to 5 requests per minute per IP, which breaks any agent loop that needs to fire queries in bursts.

ModelHub API

ModelHub is a hosted API gateway that puts 45 Chinese and global LLMs — DeepSeek V4, Qwen 3, GLM-4, Doubao, Kimi — behind a single OpenAI-compatible endpoint. You swap your base_url, keep your existing SDK, and your token bill drops. The vendor states prompts are never stored and payments run through Paddle under PCI Level 1 certification. The ceiling appears fast: no self-hosted option, no agentic tooling, no fine-tuning surface. Teams that need dedicated infrastructure or low-latency SLAs will exhaust what the service offers and contact the Enterprise tier — or leave.

Attribute	ArXiv Scholar	ModelHub API
Pricing	Free	Paid
Price	—	$15/month
Free trial	No	No
Open source	Yes	No
Has API	Yes	Yes
Self-hosted option	Yes	No
Platforms	Web API, self-hostable via GitHub	Web, CLI, API, OpenAI-compatible SDK
Released	2026-06	—
Pros	Every answer is grounded in real arXiv paper IDs, so the hallucinated-citation failure mode that breaks LLM-powered research assistants does not surface here. ML-based query routing classifies incoming questions in 1ms and selects Direct, Decompose, or HyDE paths automatically, which means complex multi-part research questions get decomposed before retrieval instead of returning a single low-precision vector match. Hybrid retrieval fuses dense BGE embeddings with BM25 sparse search and a Jina cross-encoder re-ranker, so recall stays high on both keyword-specific queries and semantically fuzzy ones — without requiring the developer to tune separate retrieval modes manually. MIT license with a public GitHub repository, so teams that need higher rate limits or want to extend the corpus can self-host and modify the full pipeline without a commercial dependency. No authentication required on the public endpoint, so an agent or prototype can start querying the live API immediately without provisioning API keys or managing credentials.	OpenAI SDK compatibility via a base_url swap, so existing codebases require no refactoring and teams avoid the integration cost of adopting a net-new client library. Per-token pricing on DeepSeek V4 Flash starting at $0.15/M tokens, which means high-volume workloads — batch summarization, large-scale code generation — that would exhaust an OpenAI budget stay economically viable. No Chinese phone number or regional payment method required, so international developers who hit identity-verification blocks on direct Chinese model APIs can provision access in minutes. Prompts are never stored and never used for model training, according to the vendor, so teams with baseline data-handling policies avoid the contractual exposure that comes with providers who retain inference data. 45 models behind one key, so switching from DeepSeek to Qwen or Doubao for a specific task is a model-name change in the request body — not a new vendor contract, new SDK, or new auth flow.
Cons	The corpus is fixed at roughly 5,600 AI engineering papers across RAG, LLMs, agents, training, and inference — any query touching adjacent domains like bioinformatics, finance, or even adjacent ML subfields returns nothing useful, and teams building cross-domain research agents have to build or integrate a separate retrieval system. The public endpoint is rate-limited to 5 requests per minute per IP; an agent running a multi-step literature review that fires sequential sub-queries will start queuing or failing at that ceiling, forcing teams to either self-host the full stack or throttle their agent's query rate to the point it defeats the purpose of automation. The autonomous agent layer described on the roadmap is marked as planned for Q4 2026 and is not shipped — teams expecting a ready-made research agent on top of this pipeline are building that orchestration layer themselves, which means this is retrieval infrastructure, not a finished agent product. The ingestion pipeline ran via Google Colab notebooks against a static pull from arXiv, so the corpus does not update continuously; a team that needs retrieval over papers published after the ingestion run must re-run the pipeline themselves on the self-hosted version — there is no documented automated refresh cadence on the public endpoint.	No self-hosted or on-premise deployment option exists. Teams under data-residency mandates that prohibit routing prompts through third-party cloud infrastructure cannot use ModelHub at all — those teams go directly to self-hostable model weights via Ollama or a private cloud deployment. The service provides chat completion inference only, with no built-in tool-use framework or agent runtime. Teams building multi-step agents that branch based on tool output must wire a separate orchestration layer — LangChain, LlamaIndex, or equivalent — on top of ModelHub, meaning they are maintaining two systems from the first agent they ship. Latency is shared-infrastructure latency with no published p99 SLA outside the Enterprise tier. Production applications where response time is a user-experience constraint — real-time voice, interactive copilots — will hit unpredictable queuing during demand spikes and have no contractual recourse short of negotiating an Enterprise deal. The full model catalog and multiple API keys are gated behind paid tiers; the free credit covers evaluation only. Teams that prototype on free credit and then need concurrent key distribution for a multi-service architecture face a hard paywall before they finish scoping the project.

Bottom line

ArXiv Scholar is free while ModelHub API is paid; ArXiv Scholar is open source. Choose based on which difference matters most for your workflow.

Frequently asked questions

What is the difference between ArXiv Scholar and ModelHub API?

ArXiv Scholar is Free and open source, while ModelHub API is Paid. Compare pricing, free trial, API, platforms, and pros/cons in the table above on AIDiveForge.

Is ArXiv Scholar better than ModelHub API?

It depends on your workflow. Use the side-by-side attributes (pricing, open source, API, self-hosted, platforms) to decide. AIDiveForge does not rank a universal winner — we publish verified facts so you can choose.