cYpher.camp

The models behind your work in cYpher.camp

Your project.
A model that
fits the job.

Choose a model for the work you want to do in AI Studio or with an AI Hire. Start with a good fit, review the result, and change models when the job calls for it.

Find my starting point Free to explore. No credits used here.
YOUR GOAL, MORE THAN ONE PATH
Make something useful.A clear brief. A result you can check.
Quick tasksKeep it light
Everyday workFind your balance
Hard problemsGo deeper
The result is what counts.Quality + total credits + your time

A higher price per token doesn’t always mean a higher price per finished task.

01 / Match the model to the work

What are we working on?

Choose the job. Then choose what matters most.

Choose a job
What matters most?

Your shortlist · Fix or write code

Try, review, adjust.

Recommendations based on provider documentation, not a measured cYpher ranking. These models power work in AI Studio and your AI Hires. Model access follows your plan; both use your plan’s shared AI credits. You review the Studio credit estimate before sending.

See plans for this work

02 / Follow the work

Credits follow the work.
Not the clock.

A task can involve much more than its final answer. Tap a step to see what happens.

Step 01

Understand the request

Your instructions, documents and conversation become input. Longer context can mean more credits.

Illustration, not a model speed comparison.

03 / Change the assumptions

A cheap attempt.
Or a cheaper result?

Try different models and attempt counts. A higher rate can still produce a lower total—if fewer attempts are needed.

Illustrative calculator · you set the attempts · no AI runs
PATH A20 credits

Each segment is one attempt. Later attempts carry more conversation.

PATH B12 credits

Each segment is one attempt. Later attempts carry more conversation.

Path B uses 8 fewer credits under these assumptions. That does not predict which model will finish your task.

See the calculation

Each attempt reads the starting context plus 2,000 tokens for every previous answer, and produces 2,000 answer or thinking tokens. Each request rounds up to at least one credit. We assume no cache hits and exclude tools and separate tool charges.

Rates use cYpher’s current Standard credit policy for short requests. DeepSeek uses peak rates; Haiku uses its short-request rate. Carbon’s introductory rate applies through November 30, 2026, then its Standard rate applies. The attempt counts are your assumptions, not measured success rates. Your actual estimate appears in AI Studio before you send.

04 / Meet the whole lineup

Different strengths.
No universal winner.

Explore all 14 models. Start with a clear job, then judge the result—not the name.

Google

Gemini 4 Hydrogen

Everyday coding and work involving several tools or steps.

A starting point based on the provider’s published guidance. Tools, instructions and reasoning settings also affect the result.

Try this model Read the provider source ↗

What the research shows

More expensive tokens.
Sometimes, less expensive work.

~½

The cost, on one coding benchmark.

Anthropic reports Fable 5.2 at High effort matching Opus 5.5’s best FrontierCode result at roughly half the cost per task.

Read the evaluation ↗
≠

Faster tokens aren’t a faster task.

In Anthropic’s latency evaluation, Sonnet finished sooner than Haiku because it wrote fewer tokens, even though Haiku generated them faster.

Read the experiment ↗

These are provider-reported examples under specific test conditions, not cYpher benchmarks or predictions for your job. Recommendations reviewed November 12, 2026.

Your next useful step

Put your choice
to work.

Your course shows you how. AI Studio is where you try the task. Your plan provides access and shared AI credits for Studio and your Hires.

Choose a model for this task See plans for this work →

Tried a model? Found something useful? Compare notes with people in Camp →