GuidesOneMux

Gemini 3 Pro Image Preview API costs for document question answering

Estimate document question-answering costs using explicit token assumptions, comparable price units, and a reproducible monthly calculation.

AI

Summary · TL;DR

An illustrative workload of 5,000 attributed document answers per day costs $1,620 over 30 days at the external source's stated standard token rates. OneMux prices for this model are unknown in the supplied catalogue snapshot, so the same estimate cannot be presented as a OneMux quote.

Key Takeaways

  • At $2 per million input tokens and $12 per million output tokens, a request with 3,000 input and 400 output tokens costs an estimated $0.0108.
  • At 5,000 requests per day, that workload costs $54 per day or $1,620 for a 30-day month, before retries and other unpriced charges.
  • The OneMux snapshot lists both official and OneMux token prices as unknown for this model; its published ratio alone does not establish a payable rate.
  • The external source contains contradictory deprecation and availability statements. Verify current access, pricing, supported inputs, and limits before deployment.
On this page (8)

The estimate at a glance

For an illustrative document question-answering workload, 5,000 requests a day at 3,000 input tokens and 400 output tokens per request would cost $54 a day, or $1,620 for a 30-day month, using the standard rates reported by Future AGI's model cost page. This is a token-only estimate, not a provider quote or a measured bill.

The decision comes with two important qualifications. First, the external page identifies Gemini 3 Pro Image Preview as an image generation model; it does not establish that the model is a good choice for attributed document answers. Second, its deprecation and availability statements conflict, so confirm that your intended provider accepts traffic before building around it.

OneMux publishes this guide and provides access to models through an OpenAI-compatible API. Its supplied catalogue snapshot does not disclose usable input or output prices for this model, so the calculation below is not a OneMux price estimate.

Price evidence and comparable units

The OneMux catalogue snapshot was captured on November 20, 2026. The external source labels its rates as verified on October 22, 2026. These are separate sources, not a reconciliation of the same price list. All known token rates below are USD per one million tokens.

Price evidenceOfficial inputOfficial outputOneMux inputOneMux outputPublished OneMux price ratio
OneMux catalogue snapshotUnknownUnknownUnknownUnknown0.1
Future AGI external model page, standard rates$2.00$12.00Not statedNot statedNot stated
Future AGI external model page, batch rates$1.00$6.00Not statedNot statedNot stated

The catalogue's ratio is reproduced as published; it does not supply the missing token prices or establish a discount that can be applied to the external rates. Unknown is not free. Check the OneMux model page and the applicable provider terms for a current quote, billing units, and any tier conditions.

The external page also lists $0.134 per image. That is a different unit from dollars per million tokens. Do not add it to a text-answer estimate without confirming when it applies and whether it overlaps with other output charges. The page advertises prompt caching but does not provide cache-read or cache-write rates for this model in the supplied summary; this estimate assumes no cache discount. Tier thresholds and batch eligibility should likewise be confirmed with the provider.

Reproduce the document Q&A calculation

Assume each request answers one question using document excerpts supplied in the prompt. The response includes an answer and source identifiers. These are planning assumptions, not measured token counts:

ComponentAssumed tokens per requestIncluded content
Input3,000Instructions, question, retrieved excerpts, and source labels
Output400Answer, citations, and any quoted supporting text
Volume5,000 requests per dayOne model call per question, with no retries

At the external page's standard rates, the arithmetic is:

  1. Input per request: 3,000 × $2.00 / 1,000,000 = $0.0060.
  2. Output per request: 400 × $12.00 / 1,000,000 = $0.0048.
  3. Total per request: $0.0060 + $0.0048 = $0.0108.
  4. Daily total: 5,000 × $0.0108 = $54.00.
  5. Thirty-day total: 30 × $54.00 = $1,620.00 for 150,000 requests.

The external calculator displays $1,644 for a month at the same daily volume. Its displayed monthly volume is 152,188 requests rather than the 150,000 used here; multiplying 152,188 by $0.0108 gives approximately $1,643.63, which explains the difference after rounding. Always compare estimates using the same month length and request count.

For the same token counts, the external page's batch rates produce $0.0030 input plus $0.0024 output, or $0.0054 per request. That is $810 for 150,000 requests, before other charges. This is only a scenario: confirm that the endpoint, model, and workload qualify for batch pricing and that the turnaround fits the application.

Budget for attribution, not just answers

Source attribution consumes tokens in both directions. Each excerpt needs a label in the input, and the answer needs enough output space to cite it. If adding labels and supporting quotations increases the average request by 500 input tokens and 100 output tokens, the additional standard-rate cost is:

  • Input: 500 × $2.00 / 1,000,000 = $0.0010 per request.
  • Output: 100 × $12.00 / 1,000,000 = $0.0012 per request.
  • Combined: $0.0022 per request, or $330 across 150,000 requests.

Under those assumptions, the 30-day estimate rises from $1,620 to $1,950. This is not a claim that attribution always adds those exact tokens; measure your own prompts and responses. Output tokens cost six times as much as input tokens at the stated standard rates, so verbose quotations can materially affect the bill.

For a practical implementation, assign each retrieved passage a stable document and page or section identifier. Ask the model to cite only identifiers included with the excerpts, and verify that returned identifiers exist before showing the answer. A valid identifier alone does not prove that the passage supports the claim: review attribution quality separately, especially when answers affect consequential decisions.

Check fit and account for missing costs

The external page reports vision input and prompt caching, but does not advertise PDF input or structured output. It lists a 131,072-token context window and a 32,768-token maximum output; the OneMux catalogue instead lists a 1M context. Treat those conflicting context figures as unresolved and verify the limit for the exact provider route and request format. Documented capabilities are not measurements of answer accuracy or citation quality.

For a text-excerpt workflow, establish how documents will be extracted and retrieved before the model call. If using page images, confirm image-input support, token accounting, and any image-related charges for the exact endpoint. The estimate excludes extraction, retrieval, storage, taxes, gateway fees, image charges, and additional tool or model calls.

Retries also matter. As a simple sensitivity check, if 5% of requests incur one additional fully billed, identical call, the standard-rate model total becomes $1,701 for 30 days. Actual retry charges depend on what the provider bills; do not assume every failed call has the same cost.

Finally, the external page says both that deprecation was scheduled for June 25, 2026 and that the provider had stopped accepting new traffic. Those statements are not enough to establish current availability for your account or route. Verify access and a migration option directly before committing to the model.

Conclusion

Use $1,620 per 30 days as a transparent baseline for the stated standard-rate scenario, not as a OneMux quote. Replace the assumed token counts with usage from a representative sample, add the costs of document processing and retries, and check current rates and model availability before making a purchasing decision.

Sources

Published price snapshot

Captured 2026-11-20T11:00:06.374Z from the OneMux published catalogue. Prices retain website billing units; this is not a live benchmark. Tiered, cached or time-based billing may differ. Unknown values are not zero.

ModelOfficial inputOfficial outputOneMux inputOneMux outputPublished price ratio
Gemini 3 Pro Image PreviewUnknownUnknownUnknownUnknown0.1

Current OneMux prices. Compare workload costs only when input/output and billing units match; see each model page for tier conditions.

FAQ

How much does Gemini 3 Pro Image Preview cost for document question answering?

At the external source's stated standard rates, 3,000 input tokens and 400 output tokens cost $0.0108 per request. At 5,000 requests per day, the token-only estimate is $1,620 over 30 days.

Does the estimate include source attribution?

Yes. The illustrative 3,000 input and 400 output tokens include source labels and citations. Actual attribution overhead depends on excerpt length, citation format, and quotations.

Is $1,620 a OneMux price quote?

No. The calculation uses rates reported by an external source. The supplied OneMux catalogue snapshot lists the model's input and output prices as unknown.

Can Gemini 3 Pro Image Preview accept PDFs directly?

The supplied external source does not advertise PDF input. Confirm support for the exact provider route, or use a separately validated document extraction workflow.

Related articles