Noemad Docs
Billing & limits
Official USD rates, bounded USDC authorization, and measured actual cost.
Integration guide · planned capabilities. Model-specific capabilities, usage rates, and x402 upto settlement require backend implementation and validation. The preview runs local samples; documented contracts do not imply live availability.
Implementation target. This guide specifies the intended product surface. The current frontend uses local samples; it does not call an inference service or collect funds. An endpoint is enabled only after its adapter, ownership, usage, and payment behavior are qualified.
Pay for the inference you use
Noemad charges a request according to the selected model’s rate and attributable usage. The user keeps tokens in their own wallet; the app does not create a prepaid balance.
Model prices are shown in USD using official sources where the exact model and billing unit can be verified. The final payment quote is denominated in USDC on Monad. An unverified model price is shown as Price not verified, never a guessed rate or a free offer.
The operator’s provider account and the user’s activity remain separate. /v1/usage must return owner-scoped usage, not the balance or logs of a shared service account.
What a quote freezes
| Term | Meaning |
|---|---|
| Owner, method, path, body hash | The exact logical request this authorization can serve |
| Provider and model IDs | The qualified configuration selected for this operation |
| USD rate snapshot and version | Input, output, cache, modality units, modifiers, and any verified offer |
| USDC conversion and rounding policy | How measured USD cost becomes token atomic units |
| Enforced usage limits | The real bounds used to calculate the maximum |
| Maximum amount | The most this one authorization permits |
| Chain, asset, merchant | Network, token contract, and exact recipient |
| Facilitator and proxy | The allowed collection path |
| Expiry | The bounded authorization window |
Use integer or fixed-point accounting. Preserve both the pre-rounding amount and the final token atomic amount. A rate change applies to future quotes; an operator must absorb an unplanned difference or reject before dispatch.
A public price table helps comparison. It is not a request-specific quote, a guaranteed availability signal, or permission to spend from a wallet.
Authorize a maximum; settle actual usage
The target x402 v2 upto mechanism authorizes a maximum and collects the actual amount after reliable usage is known. It uses the pinned Permit2/proxy mechanism from the chosen SDK. The maximum is not transferred to the merchant before inference.
- Validate the request and obtain an immutable quote.
- Review origin, model, limits, rates, maximum, network, asset, recipient, expiry, facilitator, and proxy.
- Complete any required wallet allowance, then sign the bounded authorization.
- Bind the nonce once to the owner, operation, and request hash before dispatch.
- Deliver inference and durably record attributable usage.
- Calculate the final amount and persist one settlement job.
- Verify the bound onchain receipt before marking collection settled.
The unused maximum stays in the wallet. Authorization does not reserve funds, so a balance or allowance can change before collection. An inference can succeed while collection fails; those states must remain separate.
Compute the USD cost from the right units
ordinary_input = total_input - included_cache_read
USD_cost = ordinary_input × USD_input_rate
+ cache_read × USD_cache_rate
+ output × USD_output_rate
+ other documented model-specific unitsToken rates are often quoted per one million tokens; divide the measured count by the stated unit before applying the rate. Only subtract cached input when the provider’s total-input definition already includes it. Reasoning and cache-write units follow the exact model contract.
Prefer a verified, request-correlated provider receipt where available. Otherwise calculate from validated usage and the pinned rate snapshot. A change in a shared provider account balance cannot attribute one request’s cost during concurrent use.
USD to USDC: an illustrative calculation
The following figures are a worked example, not a model price. It assumes that this example quote explicitly converts USD to USDC at 1:1 and uses a token with six decimals. Production quotes must disclose their actual conversion and rounding policy.
| Illustrative item | Calculation |
|---|---|
| Input / cached input / output | 1,201 / 800 / 203 tokens |
| Ordinary input at $1.25 per 1M | 401 × $1.25 / 1,000,000 = $0.00050125 |
| Cache read at $0.125 per 1M | 800 × $0.125 / 1,000,000 = $0.00010000 |
| Output at $5.00 per 1M | 203 × $5.00 / 1,000,000 = $0.00101500 |
| Measured USD cost | $0.00161625 |
| Quoted conversion, rounded down | 1,616 atomic units = 0.001616 USDC |
The operator absorbs the sub-atomic remainder under the planned rounding policy. It is not saved as hidden debt or a future top-up. The project plan adds no model markup; any network fee is disclosed separately along with who pays it.
Chat pricing in USD
These are official USD rates only where the exact model is verified. Input, output, cached input, cache creation, and reasoning can have different definitions. Do not substitute a nearby model’s price for an unverified identifier.
| Model | USD rate | Official source |
|---|---|---|
| GLM 5.3 | Input: $1.40 / 1M tokens Output: $4.40 / 1M tokens Cached input: $0.26 / 1M tokens | Official source ↗ Checked 2026-10-24 International API list rates. Cache storage and tool charges are separate. |
| GLM 5.3 Flash | Input: $0.15 / 1M tokens Output: $0.50 / 1M tokens Cached input: $0.03 / 1M tokens | Official source ↗ Checked 2026-10-24 International API list rates. Cache storage and tool charges are separate. |
| GLM 5.2 | Input: $1.40 / 1M tokens Output: $4.40 / 1M tokens Cached input: $0.26 / 1M tokens | Official source ↗ Checked 2026-10-24 International API list rates. Cache storage and tool charges are separate. |
| GLM 5.1 | Input: $1.40 / 1M tokens Output: $4.40 / 1M tokens Cached input: $0.26 / 1M tokens | Official source ↗ Checked 2026-10-24 International API list rates. Cache storage and tool charges are separate. |
| GLM 5.0 Turbo | Price not verified | Official source ↗ Checked 2026-10-24 This exact ID has no verified official USD price row; no price is inferred from GLM-5 or GLM-5-Turbo. |
| GLM 5V Turbo | Price not verified | Official source ↗ Checked 2026-10-24 The model has official capability documentation, but its current USD tariff was not present in the official price table checked. |
| Muse Spark 1.1 | Input: $1.25 / 1M tokens Output: $4.25 / 1M tokens Cached input: $0.15 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard mode; no long-context premium. Search grounding is billed separately. |
| Muse Spark 1.2 | Input: $1.25 / 1M tokens Output: $4.25 / 1M tokens Cached input: $0.15 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard mode; no long-context premium. Search grounding is billed separately. |
| Qwen 3.8 Max | Input: $2.00 / 1M tokens Output: $6.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Singapore international deployment, up to 1M input tokens. Standard list rates; cache, batch, and regional discounts excluded. |
| Qwen 3.7 Max | Input: $2.50 / 1M tokens Output: $7.50 / 1M tokens | Official source ↗ Checked 2026-10-24 Singapore international deployment, up to 1M input tokens. Standard list rates; cache, batch, and regional discounts excluded. |
| Qwen 3.7 Plus | Input: $0.40 / 1M tokens Output: $1.60 / 1M tokens Input above 256K context: $1.20 / 1M tokens Output above 256K context: $4.80 / 1M tokens | Official source ↗ Checked 2026-10-24 Singapore international list rates before promotions. Up to 256K input tokens: base rates; above 256K: higher tier. Cache discount excluded. |
| Kimi K3 | Input: $3.00 / 1M tokens Output: $15.00 / 1M tokens Cached input: $0.30 / 1M tokens Cache write · 5 minutes: $3.00 / 1M tokens Cache write · 1 hour: $6.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard token rates. Cache writes are billed separately by TTL; the token calculator excludes writes, tools, and taxes. |
| Kimi K2.7 | Price not verified | Official source ↗ Checked 2026-10-24 Official pricing names kimi-k2.7-code. This exact catalog ID is not confirmed as an alias; no tariff is inferred. |
| Kimi K2.6 | Input: $0.95 / 1M tokens Output: $4.00 / 1M tokens Cached input: $0.16 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard token rates. K2 access restrictions begin October 31; retirement is planned by December 2026. Taxes excluded. |
| Kimi K2.5 | Price not verified | Official source ↗ Checked 2026-10-24 This exact older version is absent from the current official pricing table. No current USD rate verified. |
| DeepSeek V4 Pro | Input: $1.32 / 1M tokens Output: $3.96 / 1M tokens Cached input: $0.044 / 1M tokens | Official source ↗ Checked 2026-10-24 Peak rates. Off-peak is 50% lower. See official UTC schedule and Chinese public-holiday exceptions. |
| DeepSeek V4 Pro 0813 | Input: $1.32 / 1M tokens Output: $3.96 / 1M tokens Cached input: $0.044 / 1M tokens | Official source ↗ Checked 2026-10-24 Official table identifies this version under API ID deepseek-v4-pro. Peak rates shown; off-peak is 50% lower. |
| DeepSeek V4 Flash | Input: $0.30 / 1M tokens Output: $1.20 / 1M tokens Cached input: $0.006 / 1M tokens | Official source ↗ Checked 2026-10-24 Accepted legacy alias now serves V4.1-Flash-1017. Peak rates shown; off-peak is 50% lower. |
| DeepSeek V4.1 Flash | Price not verified | Official source ↗ Checked 2026-10-24 Current API uses deepseek-flash for V4.1-Flash-1017. This exact version/ID mapping has no separately verified current tariff. |
| DeepSeek V4 Flash 0731 | Price not verified | Official source ↗ Checked 2026-10-24 This retired version has no current first-party USD tariff verified. The legacy unversioned alias serves a newer model. |
| MiniMax M3 | Input: $0.60 / 1M tokens Output: $2.40 / 1M tokens Cached input: $0.12 / 1M tokens Input above 512K context · list: $1.20 / 1M tokens Output above 512K context · list: $4.80 / 1M tokens Cached input above 512K context · list: $0.24 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard list rates before the displayed permanent 50% discount. Above 512K input tokens, rates double. Priority costs 1.5× standard. |
| MiniMax M2.7 | Price not verified | Official source ↗ Checked 2026-10-24 Older price references exist, but this exact model is absent from the current official pricing overview. Current USD tariff unverified. |
| Hunyuan 3 | Input: $0.132 / 1M tokens Output: $0.528 / 1M tokens Cached input: $0.033 / 1M tokens | Official source ↗ Checked 2026-10-24 Tencent TokenHub Singapore list rates. Model identifier hy3 is confirmed by the official model list. |
| Hunyuan 4 | Price not verified | Official source ↗ Checked 2026-10-24 Official model list and pricing use hy4-preview. This exact hy4 ID is not confirmed as an alias; no price is inferred. |
| Grok 4.5 | Input: $2.00 / 1M tokens Output: $6.00 / 1M tokens Cached input: $0.30 / 1M tokens | Official source ↗ Checked 2026-10-24 Global standard rates; all token rates double at 200k input tokens. Regional endpoints and tools can add charges. |
| Grok 4.6 | Input: $2.00 / 1M tokens Output: $6.00 / 1M tokens Cached input: $0.50 / 1M tokens | Official source ↗ Checked 2026-10-24 Global standard rates; all token rates double at 200k input tokens. Regional endpoints and tools can add charges. |
| GPT 5.5 | Input (up to 272K context): $5.00 / 1M tokens Output (up to 272K context): $30.00 / 1M tokens Cached input (up to 272K context): $0.50 / 1M tokens Input (over 272K context): $10.00 / 1M tokens Output (over 272K context): $45.00 / 1M tokens Cached input (over 272K context): $1.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard API, global processing, input up to 272K tokens. Meter uses short-context input/output/cache-read rates; longer prompts use the separately listed rates. Tool charges and cache writes are additional. |
| GPT 6 Astra | Input (up to 272K context): $10.00 / 1M tokens Output (up to 272K context): $50.00 / 1M tokens Cached input (up to 272K context): $1.00 / 1M tokens Input (over 272K context): $20.00 / 1M tokens Output (over 272K context): $75.00 / 1M tokens Cached input (over 272K context): $2.00 / 1M tokens 30-minute cache write (up to 272K context): $12.50 / 1M tokens 12-hour cache write (up to 272K context): $25.00 / 1M tokens 30-minute cache write (over 272K context): $25.00 / 1M tokens 12-hour cache write (over 272K context): $50.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard API, global processing, input up to 272K tokens. Meter uses short-context input/output/cache-read rates; longer prompts use the separately listed rates. Tool charges and cache writes are additional. |
| GPT 5.6 Sol | Input (up to 272K context): $4.00 / 1M tokens Output (up to 272K context): $20.00 / 1M tokens Cached input (up to 272K context): $0.40 / 1M tokens Input (over 272K context): $8.00 / 1M tokens Output (over 272K context): $30.00 / 1M tokens Cached input (over 272K context): $0.80 / 1M tokens 30-minute cache write (up to 272K context): $5.00 / 1M tokens 12-hour cache write (up to 272K context): $10.00 / 1M tokens 30-minute cache write (over 272K context): $10.00 / 1M tokens 12-hour cache write (over 272K context): $20.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard API, global processing, input up to 272K tokens. Meter uses short-context input/output/cache-read rates; longer prompts use the separately listed rates. Tool charges and cache writes are additional. Promotional rate is available at least through November 21, 2026. |
| GPT 5.6 Terra | Input (up to 272K context): $2.00 / 1M tokens Output (up to 272K context): $12.00 / 1M tokens Cached input (up to 272K context): $0.20 / 1M tokens Input (over 272K context): $4.00 / 1M tokens Output (over 272K context): $18.00 / 1M tokens Cached input (over 272K context): $0.40 / 1M tokens 30-minute cache write (up to 272K context): $2.50 / 1M tokens 12-hour cache write (up to 272K context): $5.00 / 1M tokens 30-minute cache write (over 272K context): $5.00 / 1M tokens 12-hour cache write (over 272K context): $10.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard API, global processing, input up to 272K tokens. Meter uses short-context input/output/cache-read rates; longer prompts use the separately listed rates. Tool charges and cache writes are additional. |
| GPT 5.6 Luna | Input (up to 272K context): $0.20 / 1M tokens Output (up to 272K context): $1.20 / 1M tokens Cached input (up to 272K context): $0.02 / 1M tokens Input (over 272K context): $0.40 / 1M tokens Output (over 272K context): $1.80 / 1M tokens Cached input (over 272K context): $0.04 / 1M tokens 30-minute cache write (up to 272K context): $0.25 / 1M tokens 12-hour cache write (up to 272K context): $0.50 / 1M tokens 30-minute cache write (over 272K context): $0.50 / 1M tokens 12-hour cache write (over 272K context): $1.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard API, global processing, input up to 272K tokens. Meter uses short-context input/output/cache-read rates; longer prompts use the separately listed rates. Tool charges and cache writes are additional. |
| Claude Fable 5.1 | Input: $10.00 / 1M tokens Output: $50.00 / 1M tokens Cache read: $0.25 / 1M tokens 5-minute cache write: $12.50 / 1M tokens 1-hour cache write: $20.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard first-party Claude API, global routing. Meter excludes separate cache-write and server-tool charges; batch, fast-mode and regional rates differ. |
| Claude Fable 5 | Input: $10.00 / 1M tokens Output: $50.00 / 1M tokens Cache read: $1.00 / 1M tokens 5-minute cache write: $12.50 / 1M tokens 1-hour cache write: $20.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard first-party Claude API, global routing. Meter excludes separate cache-write and server-tool charges; batch, fast-mode and regional rates differ. |
| Claude Opus 5 | Input: $5.00 / 1M tokens Output: $25.00 / 1M tokens Cache read: $0.50 / 1M tokens 5-minute cache write: $6.25 / 1M tokens 1-hour cache write: $10.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard first-party Claude API, global routing. Meter excludes separate cache-write and server-tool charges; batch, fast-mode and regional rates differ. |
| Claude Opus 4.8 | Input: $5.00 / 1M tokens Output: $25.00 / 1M tokens Cache read: $0.50 / 1M tokens 5-minute cache write: $6.25 / 1M tokens 1-hour cache write: $10.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard first-party Claude API, global routing. Meter excludes separate cache-write and server-tool charges; batch, fast-mode and regional rates differ. |
| Claude Opus 4.7 | Input: $5.00 / 1M tokens Output: $25.00 / 1M tokens Cache read: $0.50 / 1M tokens 5-minute cache write: $6.25 / 1M tokens 1-hour cache write: $10.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard first-party Claude API, global routing. Meter excludes separate cache-write and server-tool charges; batch, fast-mode and regional rates differ. |
| Claude Opus 4.6 | Input: $5.00 / 1M tokens Output: $25.00 / 1M tokens Cache read: $0.50 / 1M tokens 5-minute cache write: $6.25 / 1M tokens 1-hour cache write: $10.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard first-party Claude API, global routing. Meter excludes separate cache-write and server-tool charges; batch, fast-mode and regional rates differ. |
| Claude Sonnet 5 | Input: $2.00 / 1M tokens Output: $10.00 / 1M tokens Cache read: $0.20 / 1M tokens 5-minute cache write: $2.50 / 1M tokens 1-hour cache write: $4.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard first-party Claude API, global routing. Meter excludes separate cache-write and server-tool charges; batch, fast-mode and regional rates differ. |
| Claude Sonnet 4.6 | Input: $3.00 / 1M tokens Output: $15.00 / 1M tokens Cache read: $0.30 / 1M tokens 5-minute cache write: $3.75 / 1M tokens 1-hour cache write: $6.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard first-party Claude API, global routing. Meter excludes separate cache-write and server-tool charges; batch, fast-mode and regional rates differ. |
| Claude Haiku 4.5 | Input: $1.00 / 1M tokens Output: $5.00 / 1M tokens Cache read: $0.10 / 1M tokens 5-minute cache write: $1.25 / 1M tokens 1-hour cache write: $2.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard first-party Claude API, global routing. Meter excludes separate cache-write and server-tool charges; batch, fast-mode and regional rates differ. |
| GLM 5.3 FlashX | Input: $0.37 / 1M tokens Output: $1.25 / 1M tokens Cached input: $0.075 / 1M tokens | Official source ↗ Checked 2026-10-24 International API list rates. Cache storage and tool charges are separate. |
| Grok 4.7 | Input: $2.00 / 1M tokens Output: $6.00 / 1M tokens Cached input: $0.50 / 1M tokens | Official source ↗ Checked 2026-10-24 Global standard rates; all token rates double at 200k input tokens. Regional endpoints and tools can add charges. |
Embedding and rerank pricing in USD
Use the hosted provider’s official billing unit. An open-weight model alone does not define a hosted API price. Batches include all submitted inputs; rerank includes the query and every candidate, regardless of top_n.
| Model | USD rate | Official source |
|---|---|---|
| BGE Small EN | Price not verified | Official source ↗ Checked 2026-10-24 Open model weights. Hosting costs depend on the operator; no universal developer API list price. |
| BGE Base EN | Price not verified | Official source ↗ Checked 2026-10-24 Open model weights. Hosting costs depend on the operator; no universal developer API list price. |
| BGE Large EN | Price not verified | Official source ↗ Checked 2026-10-24 Open model weights. Hosting costs depend on the operator; no universal developer API list price. |
| BGE-M3 | Price not verified | Official source ↗ Checked 2026-10-24 Open model weights. Hosting costs depend on the operator; no universal developer API list price. |
| EmbeddingGemma 300M | Price not verified | Official source ↗ Checked 2026-10-24 Google documents this 300M open embedding model for deployment on your own hardware. No fixed first-party per-token API price is published here; hosting costs depend on deployment. |
| Qwen3 Embedding 0.6B | Price not verified | Official source ↗ Checked 2026-10-24 Open weights for self-hosting. No official hosted API USD tariff verified for this exact model; infrastructure cost varies. |
| BGE Reranker Base | Price not verified | Official source ↗ Checked 2026-10-24 Open model weights. Hosting costs depend on the operator; no universal developer API list price. |
| voyage-4-lite | Input: $0.02 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard text embedding API input-token rate after any account-specific free allowance; batch pricing differs. |
| voyage-4 | Input: $0.06 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard text embedding API input-token rate after any account-specific free allowance; batch pricing differs. |
| voyage-4-large | Input: $0.12 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard text embedding API input-token rate after any account-specific free allowance; batch pricing differs. |
| voyage-code-4 | Input: $0.12 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard text embedding API input-token rate after any account-specific free allowance; batch pricing differs. |
| voyage-finance-2 | Input: $0.12 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard text embedding API input-token rate after any account-specific free allowance; batch pricing differs. |
| voyage-law-2 | Input: $0.12 / 1M tokens | Official source ↗ Checked 2026-10-24 Standard text embedding API input-token rate after any account-specific free allowance; batch pricing differs. |
| Voyage rerank-3 | Processed input: $0.05 / 1M tokens | Official source ↗ Checked 2026-10-24 Official standard API rate for rerank-3; catalog uses a vendor-prefixed label. Billed tokens equal query tokens times document count, plus all document tokens. Account allowances and batch discounts are excluded. |
| Voyage rerank-3-lite | Processed input: $0.02 / 1M tokens | Official source ↗ Checked 2026-10-24 Official standard API rate for rerank-3-lite; catalog uses a vendor-prefixed label. Billed tokens equal query tokens times document count, plus all document tokens. Account allowances and batch discounts are excluded. |
| Voyage rerank-2.5 | Processed input: $0.05 / 1M tokens | Official source ↗ Checked 2026-10-24 Official standard API rate for rerank-2.5; catalog uses a vendor-prefixed label. Billed tokens equal query tokens times document count, plus all document tokens. Account allowances and batch discounts are excluded. |
| Voyage rerank-2.5-lite | Processed input: $0.02 / 1M tokens | Official source ↗ Checked 2026-10-24 Official standard API rate for rerank-2.5-lite; catalog uses a vendor-prefixed label. Billed tokens equal query tokens times document count, plus all document tokens. Account allowances and batch discounts are excluded. |
Media, realtime, and evaluation pricing in USD
Image count, video duration, speech characters, transcription duration, realtime sessions, and evaluation requests need separate units. A verified base rate does not establish a price for every size, quality, voice, or duration.
| Model | USD rate | Official source |
|---|---|---|
| Gpt 5.5 Image | Price not verified | Official source ↗ Checked 2026-10-24 No official API rate or documented official alias was found for this exact catalog ID. Published GPT Image prices have not been substituted. |
| Gpt 5.6 Luna Image | Price not verified | Official source ↗ Checked 2026-10-24 No official API rate or documented official alias was found for this exact catalog ID. Published GPT Image prices have not been substituted. |
| Gpt 5.6 Sol Image | Price not verified | Official source ↗ Checked 2026-10-24 No official API rate or documented official alias was found for this exact catalog ID. Published GPT Image prices have not been substituted. |
| Gpt 5.6 Terra Image | Price not verified | Official source ↗ Checked 2026-10-24 No official API rate or documented official alias was found for this exact catalog ID. Published GPT Image prices have not been substituted. |
| Grok Imagine Image | Image input: $0.002 / image Output · 1K / 2K: $0.02 / image | Official source ↗ Checked 2026-10-24 Text prompts have no separate charge. |
| Grok Imagine Image 2.0 | Image input: $0.01 / image 1K · Low: $0.04 / image 1.5K · Low: $0.05 / image 2K · Low: $0.06 / image 1K · Medium: $0.06 / image 1.5K · Medium: $0.07 / image 2K · Medium: $0.08 / image | Official source ↗ Checked 2026-10-24 Select the resolution and quality when estimating. |
| Grok Imagine Video | Image input: $0.002 / image Video input: $0.01 / second Output · 480p: $0.05 / second Output · 720p: $0.07 / second | Official source ↗ Checked 2026-10-24 Charges depend on input media and generated duration. |
| Grok Imagine Video 1.5 | Image input: $0.01 / image Output · 480p: $0.08 / second Output · 720p: $0.14 / second Output · 1080p: $0.25 / second | Official source ↗ Checked 2026-10-24 Charges depend on input media and generated duration. |
| Grok Voice TTS | Speech synthesis: $15.00 / 1M characters | Official source ↗ Checked 2026-10-24 Catalog label for the xAI TTS service; the native endpoint selects voice_id rather than this model ID. |
| Grok Voice Transcribe 1.0 | Price not verified | Official source ↗ Checked 2026-10-24 The current guide lists 2.0; pricing for version 1.0 was not confirmed. |
| Grok Voice Transcribe 2.0 | REST transcription: $0.10 / hour Streaming transcription: $0.20 / hour | Official source ↗ Checked 2026-10-24 Current transcription model confirmed in the official speech-to-text guide. |
| GPT Live | Price not verified | Official source ↗ Checked 2026-10-24 Official pricing identifies gpt-live-1; the catalog ID gpt-live has no confirmed official alias mapping. Duration-based pricing must not be substituted without that mapping. |
| Jev 1.13 | Input: $0.042 / 1M tokens Output: $0.00 / 1M tokens | Official source ↗ Checked 2026-10-24 Official version jev-1.13.0. Input only; output has no charge. |
Price the selected output settings
| Setting | What the quote must establish |
|---|---|
| Image quality and resolution | The selected model’s official output price or documented multiplier |
| Portrait or landscape dimensions | Whether size changes the billing unit or rate |
| Image edit sources | Any documented per-input or image-token cost |
| Video duration and resolution | The supported seconds, size, and corresponding rate |
| Speech voice and format | Any priced model, voice, output, or minimum-charge differences |
| Transcription length | The actual measured duration and published rounding rule |
| Realtime session | A verified session, duration, or token unit with enforceable caps |
Do not apply a generic multiplier across model families. Each modifier and any minimum must be confirmed against that exact model’s official pricing and frozen into the quote. Missing pricing means the paid route cannot be offered yet.
Inference and collection remain separate
Operation: quoted → authorized → dispatched → streaming / completed / failed / unknown
Collection: none → authorized → ready → settling → settled / zero_charge / unknown / failedA successful stream does not prove collection settled. Read GET /api/v1/payments/{quote_id} for actual cost and a verified receipt. A collection retry must not rerun inference.
Pin the actual amount before settlement retry and preserve the same authorization and nonce. Verify chain, asset, allowed proxy, payer, merchant, amount, nonce, success, and the required block state. A transfer event alone is insufficient.
Unknown usage is never charged as the maximum
Missing final usage, an ambiguous dispatch, a consumed nonce without a bound transaction, or uncertain settlement enters reconciliation. Do not automatically debit the maximum, request a fresh signature, or rerun the model.
A known zero amount closes as zero_charge without a fabricated transaction. An allowance revoke or insufficient balance can cause collection_failed; the operator must bound outstanding liability and show the real state.
Calculator and production quotes
The calculator estimates USD cost from verified official rates where available. It does not create a live USDC quote, authorize payment, or imply that the selected model is connected.
The server produces live quote terms from the qualified model and enforceable limits. Rewards and campaigns can map to explicit per-request discounts or compensation under published policy, without creating a prepaid product balance.