Every plan buys tokens at one set price. Each model has an exchange rate: the more powerful the model, the more of your plan a task draws; the lighter the model, the less. Pick a plan below and see how many tokens of each model it buys, and what a task actually costs.
How it works: your plan is a monthly bundle of skillfaber tokens at one fixed price. Each model has an exchange rate. 1,000 tokens of work on the lightest model draws about 10 from your plan; on Sonnet 5, our default, about 670; on the most powerful model, about 3,420. Same plan, same price per token; the model decides how fast you spend. And each role can run on a different model, so you match the model to the task.
PICK A PLAN · RESCALES EVERYTHING BELOW
WORKSHOP
$99/mo
5.0M skillfaber tokens included
Trying us out
STUDIO
$249/mo
14.0M skillfaber tokens included
Most teams start here
FOUNDRY
$499/mo
32.0M skillfaber tokens included
Heavier production
FACTORY
$999/mo
70.0M skillfaber tokens included
Multi-department
PROVIDER FLAGSHIPS ON AWS · WHAT 14.0M SKILLFABER TOKENS BUY ON STUDIO ($249/MO)
Your 14.0M skillfaber tokens become a different number of tokens on each model, set by its exchange rate: bar length is how many. A more favorable rate buys more, so a task costs less. Each role can run on a different model, so you match the model to the task.
Gemma 4 31B≈17× the default's work
Google
350M tokens/mo
under 1¢ per task · ~70K AI tasks
NVIDIA Nemotron 3 Super 120B A12B≈13× the default's work
NVIDIA
280M tokens/mo
under 1¢ per task · ~56K AI tasks
MiniMax M2.5≈7.4× the default's work
MiniMax
156M tokens/mo
under 1¢ per task · ~31K AI tasks
Devstral 2 123B≈4.8× the default's work
Mistral AI
100M tokens/mo
≈1¢ per task · ~20K AI tasks
DeepSeek V3.2≈4.5× the default's work
DeepSeek
93M tokens/mo
≈1¢ per task · ~19K AI tasks
Qwen3 VL 235B A22B≈3.7× the default's work
Qwen
78M tokens/mo
≈2¢ per task · ~16K AI tasks
Kimi K2.5≈3.4× the default's work
Moonshot AI
70M tokens/mo
≈2¢ per task · ~14K AI tasks
Nova Pro≈2.8× the default's work
Amazon
58M tokens/mo
≈2¢ per task · ~12K AI tasks
Grok 4.6≈2.7× the default's work
xAI
56M tokens/mo
≈2¢ per task · ~11K AI tasks
GLM 5≈2.6× the default's work
Z.AI
54M tokens/mo
≈2¢ per task · ~11K AI tasks
Claude Sonnet 5★ DEFAULT
Anthropic
21M tokens/mo
≈6¢ per task · ~4,200 AI tasks
Claude Fable 5.1≈0.2× the default's work
Anthropic
4.2M tokens/mo
≈30¢ per task · ~838 AI tasks
GPT-5.6 Sol (1M)≈0.2× the default's work
OpenAI
4.1M tokens/mo
≈30¢ per task · ~819 AI tasks
FILTER BY WHAT YOU’LL USE IT FOR
Customer-facing chat
Support, sales intake, leads
Document review
Contracts, leases, policies
Bookkeeping & data
Categorization, extraction
High-volume / batch
Classification, ETL at scale
Hard reasoning
Complex analysis, edge cases
Tokens are what you’re billed on. AI tasks are the friendlier translation: a typical task (one back-and-forth with an agent, a short document review, a step in a workflow) uses about 5,000 tokens. Sourced from OpenRouter’s 2026 LLM usage study (~6,000 prompt + ~400 completion per typical business AI request). Long-document work consumes more: analyzing a 100-page legal contract is roughly ~70,000 tokens (about 14 tasks-worth in a single pass). Quick lookups consume less. Your usage will vary; you set a hard spend cap when you sign up.
Cheaper to run than our default
on Studio · $249/mo
Stretches your bundle further. Best for high-volume work, classification, and tasks where speed beats subtlety.
Gemma 4 E2B≈67× the default's work
Google
1.4B tokens/mo
under 1¢ per task · ~280K AI tasks
high-volume simple classification
low-cost basic text tasks
+3 more
Nemotron Nano 3 30B≈34× the default's work
NVIDIA
700M tokens/mo
under 1¢ per task · ~140K AI tasks
high-volume classification
low-latency UX
+3 more
GLM 4.7 Flash≈22× the default's work
Z.AI
467M tokens/mo
under 1¢ per task · ~93K AI tasks
tool-using agents
ReACT loops
+3 more
Ministral 3 8B≈22× the default's work
Mistral AI
467M tokens/mo
under 1¢ per task · ~93K AI tasks
high-volume simple text tasks
low-latency UX
+3 more
gpt-oss-20b≈22× the default's work
OpenAI
467M tokens/mo
under 1¢ per task · ~93K AI tasks
high-volume classification
low-latency UX
+3 more
Our default: Sonnet 5
on Studio · $249/mo
What we recommend for most jobs: the capability-per-cost sweet spot of the catalog.
Claude Sonnet 5
Anthropic
21M tokens/mo
≈6¢ per task · ~4,200 AI tasks
complex reasoning and analysis
tool-using agents / ReACT loops
+3 more
Pricier to run than our default
on Studio · $249/mo
Consumes the bundle faster than Sonnet 5 but handles the hardest reasoning. Use for contracts, edge cases, or anything where accuracy is paramount.
GPT-5.6 Terra≈0.8× the default's work
OpenAI
17M tokens/mo
≈7¢ per task · ~3,350 AI tasks
tool-using agents / ReACT loops
code review and refactoring
+3 more
Claude Sonnet 4.6≈0.67× the default's work
Anthropic
14M tokens/mo
≈9¢ per task · ~2,800 AI tasks
tool-using agents / ReACT loops
long-document QA and summarization
+3 more
Claude Sonnet 4.5≈0.67× the default's work
Anthropic
14M tokens/mo
≈9¢ per task · ~2,800 AI tasks
tool-using agents
computer-use / GUI automation
+3 more
GPT-5.4≈0.64× the default's work
OpenAI
13M tokens/mo
≈9¢ per task · ~2,650 AI tasks
complex agentic and tool-using workflows
coding, code review and codegen
+3 more
GPT-5.6 Terra (1M)≈0.49× the default's work
OpenAI
10M tokens/mo
≈12¢ per task · ~2,050 AI tasks
tool-using agents / ReACT loops
long-document analysis (near-1M context)
+3 more
Ready to put a model to work?
Pick a plan on the pricing page, drop your email, and we’ll walk you through building your first role.