Qwen3-VL-235B (via DeepInfra)

OCR to markdown with Alibaba's open-weight Qwen3-VL vision model on DeepInfra's pay-per-token API; cost derived from token prices.

Using an open-weight model through DeepInfra suits teams that want cheap token-priced OCR now, with the option to self-host the same Apache 2.0 weights later. The derived cost is $1.04 per 1,000 pages, about two-thirds of it from output tokens, and the smaller Qwen3-VL-30B-A3B brings that to roughly $0.73. The drawback is compliance: DeepInfra documents no EU region or DPA.

Headquarters: 🇺🇸 United States
Parent jurisdiction: 🇺🇸 United States
EU self-serve plan: No / not documented
GDPR DPA: No / not documented
Billing: payg
Open source: No
Website: deepinfra.com
Source last checked: 2026-10-06
Pricing evidence: Modeled rates

Pricing

Published rates converted using workload assumptions. Read the notes. Token cost assumes a 1,000×1,400 px page, a 300-token prompt and 800 output tokens. Extraction quality and retries vary. How to read these estimates.

1k additional pages
$1.037

Allowances, plan limits & calculation details

Rates are shown at their recorded precision. A zero means no charge in this calculation; the vendor may charge for things we leave out. See the notes for usage assumptions and currency conversions.

Base fee / month: $0
Included pages / month (thousands): 0
1k additional pages: $1.037

Derived from DeepInfra's Qwen3-VL-235B-A22B-Instruct prices ($0.20 in / $0.88 out per M). Assumptions per page: image tokens for a ~1,000x1,400 px page, a 300-token prompt, ~800 output tokens of markdown. Qwen3-VL uses one token per 32x32 px block (patch 16, merge 2): 992x1,408 -> 31 x 44 = 1,364 + 2 = 1,366 tokens. Per 1,000 pages: input 1,000 x (1,366 + 300) = 1.666M x $0.20 = $0.333; output 0.8M x $0.88 = $0.704; total $1.04. The smaller Qwen3-VL-30B-A3B ($0.15/$0.60) costs ~$0.73. Open weights (Apache 2.0), so the same model can be self-hosted. DeepInfra documents no EU region or DPA.

Managed alternatives

At 100K pages / month. Adjust usage and requirements.

Qwen3-VL-235B (via DeepInfra): $1,244/yr estimated.

GPT-6 Luna (via OpenAI API): $719/yr estimated · $526/yr bill differenceEstimated ratesPage cost uses an assumed image-token multiplier plus 300 prompt and 800 output tokens. Confirm actual token usage.
Mistral Small 3.2 (via Scaleway Generative APIs): $845/yr estimated · $400/yr bill differenceModeled ratesToken cost assumes a 1,000×1,400 px page, a 300-token prompt and 800 output tokens. Extraction quality and retries vary.

The price difference leaves out migration costs, operating costs and any features you’d need to replace.

Self-hosted options

Operating cost is not estimated. These options are not counted as savings against managed services.

Other OCR & document parsing comparisons

Found a wrong rate or plan restriction? Report a correction with a current source.