Official rows
Pricing pages use provider-owned pricing sources, then separate input, output, cache, batch, image, video, and audio billing units.
Start with the decision you need to make. AI Jupyter connects ranking evidence, official API prices, local hardware fit, and provider changes into practical model shortlists.
Evidence engine
AI Jupyter is built around decisions a builder actually has to make: which model to test, what the API bill might become, whether a local setup will feel usable, and which provider changes deserve a fresh comparison.
Official rows
Pricing pages use provider-owned pricing sources, then separate input, output, cache, batch, image, video, and audio billing units.
Public benchmark mix
Ranking pages blend task-specific public leaderboard signals and show source coverage instead of treating missing rows as zero.
Hardware constraints
Local model pages start from RAM, VRAM, quantization, runtime, and real comfort limits before recommending a model.
Official-source radar
AI Radar checks provider-owned changelogs, docs, blogs, pricing pages, and model pages, then summarizes only decision-relevant changes.
Composite Top 5
Best AI for Coding
Composite #1
Claude Fable 5
Claude Fable 5
Anthropic
99.9
Composite
Claude Opus 4.8
Anthropic
92.7
Composite
Claude Opus 4.7 Thinking
Anthropic
89.9
Composite
Claude Opus 4.8 Thinking
Anthropic
89.9
Composite
Claude Opus 4.7
Anthropic
89.9
Composite
Best AI for Writing
Composite #1
Claude Fable 5
Claude Fable 5
Anthropic
97.4
Composite
Gemini 3 Pro
97.1
Composite
Claude Opus 4.7 Thinking
Anthropic
96.7
Composite
Claude Opus 4.6 Thinking
Anthropic
96.2
Composite
Claude Opus 4.7
Anthropic
96.1
Composite
Best AI for Math
Composite #1
Claude Fable 5
Claude Fable 5
Anthropic
95.9
Composite
Gemini 3.5 Flash
91.2
Composite
Claude Opus 4.6 Thinking
Anthropic
91.2
Composite
GPT-5.4 High
OpenAI
90.5
Composite
Claude Opus 4.6
Anthropic
90.5
Composite
Best AI for Image Generation
Composite #1
GPT Image 2
GPT Image 2
OpenAI
100
Composite
Nano Banana 2 / Gemini 3.1 Flash Image Preview
92.4
Composite
GPT Image 1.5 High Fidelity
OpenAI
92.3
Composite
HiDream-O1-Image-1.5
HiDream
91.3
Composite
Reve 2.0
Reve
90.2
Composite
Best AI for Writing Essays
Composite #1
Claude Fable 5
Claude Fable 5
Anthropic
96.1
Composite
Gemini 3 Pro
95.9
Composite
Claude Opus 4.6 Thinking
Anthropic
95.3
Composite
Claude Opus 4.7 Thinking
Anthropic
94.9
Composite
Claude Opus 4.7
Anthropic
94.3
Composite
Categories
4
Text
Chat, reasoning, coding, agents, RAG, extraction, and long-context text workloads.
Cost per 1,000 runs
Chat, coding, agents, RAG
Images
Text-to-image, image editing, native image generation, and production image pipelines.
Cost per image
Generation, editing, product media
Video
Text-to-video, image-to-video, video-to-video, editing, reframe, and ad-creative workloads.
Cost per clip
Ads, social clips, creative testing
Audio
Text-to-speech, speech-to-text, realtime voice agents, translation, dubbing, and sound workflows.
Cost by audio unit
TTS, STT, realtime voice
Hardware guides
24
8 GB RAM laptop
Practical local LLM picks for 8GB laptops and CPU-only machines.
16 GB RAM machine
Balanced local LLMs for 16 GB laptops and MacBooks.
32 GB RAM desktop
Stronger local LLMs for 32 GB RAM systems.
RTX 5090 workstation
Top-end single-GPU local LLM picks for 32 GB VRAM RTX 5090 builds.
RTX 5080 workstation
Practical local LLM picks for 16 GB VRAM RTX 5080 systems.
RTX 5070 Ti workstation
Balanced 16 GB VRAM local LLM picks for RTX 5070 Ti machines.
RTX 5070 workstation
12 GB VRAM local LLM picks for RTX 5070 desktop builds.
RTX 5060 Ti 16GB workstation
16 GB VRAM local LLM picks for RTX 5060 Ti 16GB systems.
RTX 5060 Ti 8GB workstation
8 GB VRAM local LLM picks for RTX 5060 Ti 8GB systems.
RTX 5060 workstation
8 GB VRAM local LLM picks for RTX 5060 systems.
RTX 4090 workstation
High-performance local LLMs for 24 GB VRAM RTX 4090 builds.
RTX 4080 workstation
16 GB VRAM local LLM picks for RTX 4080 and RTX 4080 Super systems.
RTX 4070 Ti Super workstation
16 GB VRAM local LLM picks for RTX 4070 Ti Super desktops.
RTX 4070 Ti workstation
12 GB VRAM local LLM picks for RTX 4070 Ti builds.
RTX 4070 workstation
12 GB VRAM local LLM picks for RTX 4070 and RTX 4070 Super systems.
RTX 4060 Ti 16GB workstation
16 GB VRAM local LLM picks for RTX 4060 Ti 16GB systems.
RTX 4060 Ti 8GB workstation
8 GB VRAM local LLM picks for RTX 4060 Ti 8GB systems.
RTX 4060 workstation
8 GB VRAM local LLM picks for RTX 4060 systems.
RTX 3090 workstation
24 GB VRAM local LLM picks for RTX 3090 workstations.
RTX 3080 workstation
10 GB VRAM local LLM picks for RTX 3080 systems.
RTX 3070 workstation
8 GB VRAM local LLM picks for RTX 3070 systems.
RTX 3060 Ti workstation
8 GB VRAM local LLM picks for RTX 3060 Ti systems.
RTX 3060 workstation
12 GB VRAM local LLM picks for RTX 3060 systems.
Apple Silicon MacBook
MacBook-friendly local LLMs for Apple Silicon unified memory.
What changed
The core paths are AI Rankings, API Prices, Local Models, LLM Tests, and AI Radar. Each path links into methodology, official sources, real-world tests, or pricing data.
Review methodologyResearch paths
FAQ
AI Jupyter is a model-decision site for comparing AI rankings, official API prices, local LLM hardware fit, and official provider changes.
The homepage points to auditable rankings, pricing calculators, hands-on local tests, and official-source radar items instead of a generic news feed.
Use API Prices when you need production cost estimates that include retries, cache hits, batch jobs, long prompts, and output-token-heavy workloads.
Use Local Models when you want to run a model on your own machine, then narrow the choice by 8GB RAM, 16GB RAM, RTX 4090, MacBook, or real-world tests.