Pricing reviewed September 7, 2026Review dates and sources appear on each record.
Illustrated language model engine connecting prompts, context, tools, and outputs
Use the cloud whenYou want the frontier now
Run local whenControl matters more
Model catalogFamilies, variants, modalities, and every major place they run.Search catalog →Platform mapDirect APIs, clouds, gateways, local runtimes, and creative suites.Compare access →Frontier modelsWhat is current, what it costs, and how to get access.Open guide →Local AI labDGX Spark, RTX, Mac, model sizes, and software stacks.Open guide →Creative AIComfyUI, image, video, audio, and real production workflows.Open guide →Things to buildUseful projects for your work, home, team, and creative practice.Open library →

Current frontier snapshot

A short list you can act on.

These are sensible starting points—not permanent winners. Test the best two candidates on work you actually care about.

OpenAIFrontier

GPT-6 Astra

Complex reasoning, coding, and sustained professional work

API list price
$10 in · $50 out
Get access
OpenAI API
Official details
AnthropicFrontier

Claude Fable 5.1

Long-running agents and difficult reasoning

API list price
$10 in · $50 out
Get access
Anthropic API
Official details
GoogleFrontier

Gemini 3.8 Flash

Fast multimodal work and grounded research

API list price
$0.75 in · $3.75 out
Get access
Google API
Official details
xAIFrontier

Grok 4.6

Coding, agents, and current web/X research

API list price
$2 in · $6 out
Get access
xAI API
Official details

Per 1M text tokens, standard short-context API rates reviewed September 7, 2026. Gemini Flash includes a dated promotion. See the comparison for long-context tiers, promotion dates, and additional charges.

Compare a workload and calculate its cost

Cloud or local?

Use both.
Give each the right job.

Cloud models win when you need the newest intelligence, managed tools, and no hardware maintenance. Local models win when your data should stay close, use is frequent, or you need deep control over the stack.

Compare local setups

Choose cloud first

  • You use AI occasionally
  • You need the strongest models
  • You want web, files, and tools already connected
  • You do not want to manage drivers or model files

Choose local first

  • You work with sensitive source material
  • You generate enough media to resent per-use fees
  • You need repeatable, inspectable workflows
  • You enjoy owning the system

Follow a complete tutorial

Start with a result you can check.

Real things to do

Projects that earn their keep.

The best first project is useful even when the model is imperfect, small enough to inspect, and based on information you understand.

Research brief

Give it a question, a source list, and a claim table. Require citations and a section named ‘What is still uncertain?’

Private document search

Run an open model locally over contracts, notes, manuals, or journals without uploading the originals to a consumer chatbot.

Coding partner

Ask for a plan, a test that reproduces the problem, the smallest fix, and a final review—not just a block of code.

Creative production line

Turn a brief into keyframes, generate variations in ComfyUI, animate the approved frames, then create music and effects.

Meeting to action

Transcribe audio, separate decisions from discussion, draft the follow-up, and route tasks to the right owner.

Personal operating manual

Build a searchable record of your preferences, recurring decisions, checklists, and examples of work you consider good.

Browse the full ideas library and starter plans

A sane first weekend

One task. Two models. Five examples.

PickChoose a task you already do and can judge.

TestRun five representative examples through two models.

MeasureTrack quality, time, cost, and how much correction was needed.

KeepSave the prompt, examples, and failure cases as your first evaluation set.