Builders choosing a model stack
Use AI Jupyter to find the few models worth testing on your own prompts, repositories, documents, latency target, and budget.
About
AI Jupyter starts with the machine, not the model hype.
The local LLM pages start from one plain question: what happened when a model ran on a real machine. Hosted model pages and API prices can narrow the next test, but the core record is the part that can be checked later: machine, settings, speed, memory pressure, screenshots, and limits.
Why this site exists
A model list is easy to publish. A useful record is harder: it has to show the machine, the setup, the model tag, the speed, the memory behavior, and the moment the model stops feeling usable. AI Jupyter keeps model pages and price rows, but those come after the page has answered what to try next on a real machine or project.
A model page is a starting point, not a final answer. Public source rows, official prices, local hardware constraints, and test results all age quickly. That is why the site links back to sources, shows update dates, and separates model quality from the API bill or whether the model will run locally.
Who AI Jupyter is for
Teams estimating API spend
Use the pricing pages to turn official provider rows into workload assumptions such as retries, cache hits, output length, and batch eligibility.
Local AI users checking hardware fit
Use the local model pages and test records to decide what to try on 8GB RAM, 16GB RAM, 32GB RAM, RTX 4090, or Apple Silicon before downloading oversized models.
What stays with the record
The useful part is the record a reader can check: the machine, setup, screenshots, logs, prices, and the next test to run before trusting a model.
Start from the run
Pages start from a concrete question: what happened when this model ran on this machine or workload?
Keep page types separate
Hardware tests, public model signals, API prices, and local model fit stay separate so one score does not hide the tradeoffs.
Leave a retest path
Core pages point to the next prompt, price check, or local run a reader can do before trusting the result.
Clear limits
AI Jupyter states where tests, model scores, prices, and local model picks can be incomplete, stale, or wrong for a specific use case.
Local LLM test records
Model runs on fixed hardware, with screenshots, measured speed, memory pressure, and the point where things get too slow.
Local LLM fit
Hardware-aware guidance for RAM, VRAM, Ollama, LM Studio, llama.cpp, and open-weight model choices.
Hosted model pages
Task pages for coding, writing, math, image generation, and essays, used to pick candidates for your own tests.
Official API prices
Provider price rows, workload calculators, and source checks for text, image, video, and audio APIs.
How the pages are maintained
Source-first editing
Model pages prefer public leaderboards and task-specific evaluations. API price pages use official provider sources. Local test pages keep the hardware setup, screenshots, logs, and test notes close to the model pick.
Human review before conclusions
AI can help collect and summarize information, but the site is edited around human checks: source quality, dates, usefulness to builders, and whether a claim should be softened or removed.
What AI Jupyter does not claim
- AI Jupyter does not claim that one score or one test run decides the best model.
- AI Jupyter does not sell API access, route traffic, or resell provider credits.
- AI Jupyter does not replace testing on your own prompts, budget, latency, privacy rules, or quality bar.
Editor and corrections
AI Jupyter is edited by Guozhen, an independent publisher focused on AI model comparison, public model-signal sources, official API pricing, and practical local LLM test records. Correction requests, source suggestions, and disclosure questions can be sent to james.jupyter@gmail.com.
AI Jupyter is not affiliated with the model providers, public model-signal sources, or test tools referenced in the pages unless explicitly stated. Company names, product names, model names, and logos are used for identification and comparison.