Product

One buying decision. One flagship model.

Toolkit keeps the public story narrow: lower published token pricing, domestic paid routing, and proof that makes generic incumbents feel overpriced.

Text inference$0.25 / 1M input
Tool calls$0.015 / call
Image generation$0.08 / image

Why Switch

Lower cost. Domestic control. Proof that survives production.

Three reasons to move: published pricing that undercuts incumbents, US-hosted paid inference, and benchmarks you can reproduce.

Cost

The price should make the switch obvious.

Toolkit publishes rates the way the market already thinks. Cheap should be visible before the buyer opens a spreadsheet.

  • Subscriptions under $10/month
  • Tool calls 40% cheaper than OpenAI
  • Image generation from $0.08 per image
Control

Domestic is an operating rule, not a vibe.

Paid traffic stays on approved US-hosted infrastructure or fails closed. Privacy language has to map to a request path.

  • US-hosted inference for paid traffic
  • Fail closed if no eligible domestic region is healthy
  • Retention and routing policies map to the account
Proof

Cheap still has to survive comparison.

Cheap only matters if the model still ships useful work. The benchmark wall and gallery exist to prove that.

  • Commercial prompts instead of toy demos
  • Mobile, tablet, and desktop review
  • Methodology published beside the boast

Output gallery

Real page categories. Real prompts.

SaaS, restaurants, local services, dashboards, ecommerce, and real estate. The same categories buyers actually build.

Developer

One request. OpenAI-compatible.

Drop in the endpoint, pass your key, and ship. Same SDK, same format, lower cost.

OpenAI-compatible requesttoolkit-llm-base
curl https://toolkit-api.roseyballs.workers.dev/v1/chat/completions \
  -H "Authorization: Bearer tk_live_xxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "toolkit-llm-base",
    "messages": [
      {
        "role": "user",
        "content": "Build a SaaS pricing page with strong mobile breakpoints."
      }
    ]
  }'
Model spec
BackboneGLM-5
PositioningLightweight · Free at source
Context length200K tokens
Max output128K tokens
ModalityText → Text + Image
CapabilitiesThinking · Streaming · Tools · Image Gen
InferenceUS · Canada · EU
CompatibilityOpenAI SDK