Qwen3.7 Plus API cost for document question answering
Estimate Qwen3.7 Plus API costs for document question answering with source attribution, using explicit token assumptions and reproducible arithmetic.
Summary · TL;DR
An illustrative 150,000-request monthly document Q&A workload costs $276 at the supplied standard input and output rates, or $179.40 at OneMux's published lower-tier rates, before caching and other production costs.
Key Takeaways
- At 3,000 input and 400 output tokens per request, the illustrative OneMux lower-tier model charge is $0.001196 per request.
- Source excerpts and citation-bearing answers both consume tokens; measure them in a representative workload rather than budgeting only for the user's question.
- Cache rates, eligibility, context limits, and tier conditions differ by offering and should be confirmed before forecasting a bill.
- The estimate excludes retrieval infrastructure, document preparation, retries, and additional model calls.
On this page (8)
The short answer
For an illustrative document question-answering workload of 150,000 requests per month, with 3,000 input tokens and 400 output tokens per request, the estimated model charge is $276 at the supplied standard rates or $179.40 at OneMux's published lower-tier rates. These are token charges, not a complete cost of running a document Q&A service.
OneMux, the publisher of this article, offers model access through an OpenAI-compatible API. The comparison below uses its published catalogue snapshot captured on November 12, 2026, and the Fireworks AI listing described by Future AGI's Qwen3.7 Plus calculator. The calculator reports input and output prices of $0.40 and $1.60 per million tokens, respectively. Check current terms before committing a budget.
Price inputs and billing units
The table uses USD per 1 million tokens. The official columns are the rates labelled official in the supplied OneMux catalogue; the Fireworks listing independently reports the same standard input and output amounts. That does not establish that the offerings have identical limits, caching rules, or other terms.
| Qwen3.7 Plus context tier | Official input | Official output | OneMux input | OneMux output | Published OneMux-to-official price ratio |
|---|---|---|---|---|---|
| ≤256K | $0.40 | $1.60 | $0.26 | $1.04 | 0.65 |
| >256K | $1.20 | $4.80 | $0.78 | $3.12 | 0.65 |
A ratio of 0.65 means the listed OneMux rate is 65% of the corresponding listed official rate. Confirm the applicable tier and its conditions on the Qwen3.7 Plus model page; do not assume that a request near the boundary is billed in a particular tier without checking. The examples below assume the ≤256K tier throughout.
The source listings also differ on context: Future AGI reports a 262,144-token context window and up to 65,536 output tokens for the Fireworks offering, while the supplied OneMux catalogue lists a 1M context. Treat these as offering-specific listings, not interchangeable limits. Verify the selected endpoint's current limits, including how input and output share the available context, before sending long documents.
Calculate a monthly document Q&A estimate
Assume a retrieval step has already selected relevant passages. Each model request includes the user's question, instructions, passage identifiers, and excerpts totalling 3,000 input tokens. The answer and its source references total 400 output tokens. There is one model call per question and no cache use or retries.
At 5,000 questions per day for a 30-day month:
- Requests: 5,000 × 30 = 150,000.
- Input: 150,000 × 3,000 = 450 million tokens.
- Output: 150,000 × 400 = 60 million tokens.
| Charge | Standard-rate calculation | OneMux lower-tier calculation |
|---|---|---|
| Input | 450 × $0.40 = $180.00 | 450 × $0.26 = $117.00 |
| Output | 60 × $1.60 = $96.00 | 60 × $1.04 = $62.40 |
| Total per 30-day month | $276.00 | $179.40 |
| Total per request | $0.00184 | $0.001196 |
| Total per 1,000 requests | $1.84 | $1.196 |
The result is sensitive to how much evidence you send and how much the model returns. For example, adding 1,000 input tokens to every request adds 150 million monthly input tokens: $60 at the standard input rate or $39 at the OneMux lower-tier input rate. Adding 100 output tokens per answer adds 15 million monthly output tokens: $24 or $15.60, respectively.
The estimate excludes document ingestion, OCR, embeddings, search or vector storage, application hosting, taxes, and any gateway fees. It also excludes failed or retried calls and additional model calls for query rewriting, citation checking, or answer revision. Price those separately; an unknown cost is not a zero cost.
Budget for source attribution
A useful answer should identify the evidence a reader can inspect, not merely append a plausible document title. For an illustrative implementation:
- Assign each retrieved passage a stable document ID and location, such as a page or section number. Include those identifiers with the excerpts sent to the model.
- Ask for each material claim to cite the supplied passage IDs, and instruct the model to say when the excerpts do not support an answer.
- Check that returned IDs exist in the retrieved set. Review whether the cited passages actually support the claims; a valid ID alone does not prove attribution is correct.
- Record input and output usage for representative questions, including long answers and no-answer cases. Recalculate using the observed distribution as well as the average.
Passage labels, citation instructions, and references all count toward the example's token totals. If your application makes a separate model call to verify citations, add that call's input and output charges rather than hiding it inside the original 400-token output assumption. The supplied sources do not establish citation accuracy or measured performance for this workflow, so evaluate those on your own documents before choosing a production budget.
Treat cache savings as conditional
The supplied OneMux catalogue lists lower-tier cache-read and cache-write prices of $0.026 and $0.325 per million tokens, respectively. Its higher-tier figures are $0.078 and $0.975. Future AGI's Fireworks calculator lists cached input at $0.080 per million, while its capability section says prompt caching is not advertised. These figures and descriptions should not be treated as proof that the same cache mechanism or eligibility applies across offerings.
As a sensitivity example only, suppose every request has 2,400 eligible input tokens billed as OneMux lower-tier cache reads, 600 ordinary input tokens, and 400 output tokens. Assuming the cache is already populated and no writes are charged during the measured period:
- Cached input: 360 million tokens × $0.026 = $9.36.
- Ordinary input: 90 million tokens × $0.26 = $23.40.
- Output: 60 million tokens × $1.04 = $62.40.
- Total: $95.16 per 150,000 requests.
This is not a promised cache rate or a complete caching forecast. Repeated document prefixes may be eligible only under specific provider rules, and cache creation, expiry, or misses can change the result. Confirm how writes are billed and include them in a real estimate. Do not apply a cache-read price to all retrieved passages simply because they come from the same document collection.
Before approving the budget
Use the no-cache estimate as a transparent baseline, then replace the illustrative token counts with observed usage. Check the current endpoint's context and tier rules, reserve budget for retries and attribution checks, and review citation quality separately from price. OneMux's current pricing page is a starting point for checking whether the published rates have changed.
Sources
- Future AGI: Qwen3.7 Plus on Fireworks AI — reported model rates, limits, and capability listing.
- OneMux: Qwen3.7 Plus — published model pricing and context-tier information; catalogue snapshot used here was captured November 12, 2026.
Published price snapshot
Captured 2026-11-12T22:30:06.019Z 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.
| Model | Official input | Official output | OneMux input | OneMux output | Published price ratio |
|---|---|---|---|---|---|
| Qwen3.7 Plus | $0.4 | $1.6 | $0.26 | $1.04 | 0.65 |
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 Qwen3.7 Plus cost for 150,000 document questions?
With 3,000 input and 400 output tokens per question, the illustrative 30-day model charge is $276 at the supplied standard rates or $179.40 at OneMux's published lower-tier rates. Retrieval, retries, caching charges, and other operating costs are excluded.
Do source citations increase API costs?
They can. Passage identifiers and citation instructions add input tokens, while references in the answer add output tokens. A separate model-based citation check adds another call.
Can I assume all repeated document text receives the cache-read rate?
No. Cache eligibility, writes, expiry, and billing depend on the selected offering and its terms. Measure actual cache usage and include cache-write charges where applicable.
Related articles
Guides
GPT 6.2 Astra API cost estimate for document question answering
Estimate GPT 6.2 Astra document question-answering costs using explicit token assumptions, context tiers, cache rates, and reproducible arithmetic.
Guides
Estimate Claude Fable 5.2 API Costs for Document Q&A
A reproducible way to estimate Claude Fable 5.2 costs for document question answering, with explicit token assumptions, source-attribution guidance, and dated price figures.
Guides
Estimate GPT 6.1 Terra API Costs for Document Q&A
A reproducible way to estimate GPT 6.1 Terra costs for document question answering with source attribution, including cache assumptions and the long-context tier.
Guides
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.