Monthly AI token spend and a cross-model comparison from a usage profile and a dated price table: per-model monthly cost, cost per request, the input / cached-input / cache-write / output breakdown, and a cheapest-first ranking. Prices are an input, never kernel constants: the default prices below are a dated snapshot of the official provider price pages.
AI SpendDated SnapshotInteger MicrosZero PIIClient-side only
🔒 All inputs are processed locally in your browser. No data is transmitted. Do not enter real personal data; use synthetic or anonymised inputs only.
Scope & reliance
Pricing date as_of is an input; the node reads no clock. A price entry applies when effective_from ≤ as_of ≤ effective_through (null bounds are open); two applicable entries for a compared model are an AMBIGUOUS_PRICE refusal, and a missing block or component price is a refusal naming the model and the component. Batch and long-context shares split each component's tokens into disjoint slices priced from the matching block (the batch block, the long-context block, or the combined batch+long-context block when the provider lists one; otherwise NO_PRICE_FOR_COMBINATION). The cost of T tokens at price P is floor(T/1e6)·P + floor((T mod 1e6)·P/1e6), overflow-safe under the validated caps (monthly tokens per component at most 10¹²; prices at most 10⁹). The estimate covers text tokens only: taxes and regional uplift, image, audio and video tokens, tool-use surcharges, tiered volume discounts, and subscription or committed-spend deals sit outside the estimate, and this node ships compute_proof_ready: ready; see the proof-status note below the results.
Usage profile, price table and comparison set
Edit any field and run. The sample prices the 2026-10-04 snapshot below at a 60%-cached usage profile across one model per provider.
Result
Priced snapshot
Proof status
This node ships compute_proof_ready: ready: ZK-proven. The shard pins a zkVM execution receipt (risc0/groth16-bn254) whose seal commits to this kernel's source digest, and the result above is hash-committed in the receipt journal and recomputable from the same inputs.
Execution Hash & §4 Artifact
SHA-256 execution hash (JCS canonical, RFC 8785):
Ask your agent
Copy this paragraph into Claude, OpenClaw, or any MCP-aware agent to run this exact tool, with this sample, and verify the artifact.
Run the AINumbers MCP tool `estimate_ai_token_spend`. Task: Monthly AI token spend estimator and cross-model comparator: per-model monthly cost, cost per request, the component breakdown, and a cheapest-first ranking, priced from a dated price table by an as_of date.
Call it with arguments: {"policy_parameters":{"as_of":"2026-10-01","usage":{"requests_per_month":100000,"input_tokens_per_request":3000,"cached_input_share_bp":0,"cache_write_tokens_per_request":0,"cache_write_ttl":"5m","output_tokens_per_request":600,"batch_share_bp":0,"long_context_share_bp":0},"price_table":{"snapshot_verified_on":"2026-10-01","currency":"USD","unit":"usd_micros_per_million_tokens","models":[{"provider":"anthropic","model":"claude-haiku-4-5","tier":"standard","effective_from":null,"effective_through":null,"input":1000000,"cached_input":100000,"cache_write_5m":1250000,"cache_write_1h":2000000,"output":5000000,"long_context":null,"batch":{"input":500000,"output":2500000},"source_sha256":"0101010101010101010101010101010101010101010101010101010101010101"}]},"compare":["claude-haiku-4-5"]}}
Verify before trusting: call `verify_execution_hash` on mcp.ainumbers.co (https://mcp.ainumbers.co/mcp) with the parameter `claimed_hash` set to the returned `execution_hash`, passing the full artifact the run returned (the object containing `policy_parameters` + `output_payload` + `execution_hash`; equivalently `policy_parameters` + `output_payload` with `claimed_hash`), not the bare hash string, or re-run the in-page WebMCP tool `estimate_ai_token_spend`.
Return the ledger link https://ledger.ainumbers.co/ so a human can re-verify without contacting us.
PII rule: All inputs are processed locally in your browser. No data is transmitted. Do not enter real personal data — use synthetic or anonymised inputs only.
Over MCP, the run happens on the AINumbers worker. No AINumbers server stores your inputs.
Open the tool with the sample prefilled: https://ainumbers.co/chaingraph/art-704-ai-token-spend.html#p=v1.H4sIAAAAAAAA_wECA_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