What do input and output prices mean?
Input price is the cost of tokens sent to a model. Output price is the cost of tokens generated by the model. InferenceHunt displays both in US dollars per 1 million tokens.
InferenceHunt / Methodology
How InferenceHunt compares AI model prices
InferenceHunt compares the latest successful public price observations for the same AI model across tracked venues. Prices are shown in US dollars per 1 million tokens. Missing prices are never estimated.
Current coverage
InferenceHunt currently has pricing data for 32 models across 13 venues, with the latest successful observation collected 2026-10-07T14:34:39.841Z.
Primary sources
Prices are collected from public venue APIs and retained as source-linked observations.
- OpenRouterPublic model catalog and pricing API
- Vercel AI GatewayPublic model catalog and pricing API
- Surplus IntelligencePublic market API
- Cloudflare AI GatewayAuthenticated model catalog and published Workers AI pricing
- OpenCode ZenPublic model availability API and reviewed official pricing
- OrbioLive fee-inclusive quote for a 20 USD face-value credit purchase, applied to public catalog token rates
- AnthropicReviewed official first-party standard pricing
- OpenAIReviewed official first-party standard pricing
- Google Gemini APIReviewed official first-party standard pricing
- DeepSeekReviewed official peak pricing; off-peak rates retained as conditions
- Z.aiReviewed official first-party standard pricing
- Xiaomi MiMo APIReviewed official overseas pay-as-you-go pricing
Comparison rules
STEP 1
Match the exact model
Venue listings are mapped to one canonical model. Similar model families or variants are not combined.
STEP 2
Keep offers separate
Each venue and processing mode remains a distinct offer. Cached-token prices are not substituted for standard input prices.
STEP 3
Compare like-for-like units
Input and output prices are normalized to USD per 1 million tokens before comparison.
STEP 4
Use the latest successful observation
Current means collected within 24 hours. Older data is labeled stale and unavailable data is shown as N/A.
STEP 5
Infer discounts conservatively
A venue must be at least 20% lower for both input and output than the first-party lab, or than three non-marketplace venues publishing the same prices when no lab rate is tracked. Results are labeled “discounted” as an inference from observed prices, not as an advertised promotion.
STEP 6
Include executable credit costs
For Orbio, nominal model rates are multiplied by the total cash cost of its live 20 USD retail credit quote, including fees. The benchmark amount and liquidity dependence are disclosed with every offer.
Pricing questions
How does InferenceHunt calculate the cheapest price?
InferenceHunt compares the latest successful price observation for the same canonical model at each tracked venue. The lowest input and output prices are selected independently, so they can come from different venues.
How current is the pricing data?
A successful observation collected within the past 24 hours is current. Older observations are labeled stale, and missing values are shown as N/A rather than estimated.
Does free mean permanently free?
No. Free means the latest observed input and output prices are both zero. Promotions and venue pricing can change, so the observation time and source should be checked before use.
Why can the same model have different prices across venues?
Venues can use different providers, routing, processing modes, promotions, and margins. InferenceHunt keeps each observed offer separate so those differences remain visible.
How does InferenceHunt identify discounted models?
A model lab’s current first-party price is the authoritative baseline. When no tracked lab price is available, at least three non-marketplace venues must publish the same input and output prices for the same processing mode. A venue is discounted only when both prices are at least 20% lower than that baseline. This is an inference from observed prices, not a claim about an advertised discount.
How are Orbio prices calculated?
Orbio charges nominal model rates against prepaid credits. InferenceHunt applies Orbio’s live fee-inclusive quote for a 20 USD face-value credit purchase—the default amount shown by Orbio—to those token rates. The displayed effective price varies with purchase size and available order-book liquidity.