Pricing

All features are priced using credits, which are billed per page (or minute for audio). Credits vary by parsing mode, model, and whether files are cached.

---

Credit Conversion Rate

$1.25 per 1,000 credits.

Parsing

v2 (Tiers)

In the v2 API, parsing uses a simplified tier-based system. Choose a tier and the platform handles model selection automatically.

Tier	Credits per Page
Fast	1
Cost-effective	3
Agentic	10
Agentic Plus	45

Additional Configs:

• Layout extraction: +3 credits per page (can be added to any tier)
• Enriched forms output (Beta): +10 credits per page containing a form (details; not available on Fast)
• Spreadsheet: 1 credit per sheet
• Audio: 3 credits per minute
• Markdown (.md): the file itself bills as a single Fast page (1 credit) regardless of length, plus 1 page at the job’s tier rate for each successfully extracted inline image (base64 data URI or remote URL; relative/local paths are never extracted). Fast-tier jobs don’t extract inline images, so a .md file on Fast always bills exactly 1 credit.

v1 (Modes)

In the v1 API, you select a parse mode and model directly.

Category	Mode	Model	Credits per Page
Recommended Settings	Cost-effective	-	3
	Agentic	-	10
	Agentic Plus	-	45
Presets	Invoice	-	90
	Scientific papers	-	90
	Technical documentation	-	90
	Forms	-	90
Modes	Parse without AI	-	1
	Parse page with LLM	-	3
	Parse page with LVM	anthropic-sonnet-3.5 (deprecated)	60
		anthropic-sonnet-3.7	60
		anthropic-sonnet-4.0	60
		anthropic-sonnet-4.5	60
		anthropic-haiku-4.5 (preview)	30
		openai-gpt-4o-mini	15
		openai-gpt-4o	30
		openai-gpt-4-1-nano	15
		openai-gpt-4-1-mini	20
		openai-gpt-4-1	30
		openai-gpt-5-nano	10
		openai-gpt-5-mini	30
		openai-gpt-5	150
		gemini-2.0-flash	6
		gemini-2.5-flash	25
		gemini-2.5-pro	60
		Custom Azure Model	1
		Use your own API key	1
	Parse page with Layout Agent	-	45
	Parse page with Agent	anthropic-sonnet-3.5 (deprecated)	45
		anthropic-sonnet-3.7	90
		anthropic-sonnet-4.0	90
		anthropic-sonnet-4.5	90
		anthropic-haiku-4.5 (preview)	45
		openai-gpt-4-1-mini	10
		openai-gpt-4-1	45
		openai-gpt-5-nano	10
		openai-gpt-5-mini	45
		openai-gpt-5	90
		gemini-2.0-flash	10
		gemini-2.5-flash	10
		gemini-2.5-pro	45
	Auto Mode	-	3–45
	Parse document with LLM	-	30
	Parse document with Agent	anthropic-sonnet-3.7	90
		anthropic-sonnet-4.0	90
		anthropic-sonnet-4.5	90
Other Options	Structure Output	-	3
File Type Modes	Spreadsheet	-	1 per sheet
	Audio	-	3 per minute
Legacy Modes	Continuous Mode	-	30

Additional Configs:

• Layout extraction: +3 credits per page (can be added to any mode)
• Markdown (.md): the file itself bills as a single page at the selected mode’s rate regardless of length, plus 1 page at the same rate for each successfully extracted inline image (base64 data URI or remote URL; relative/local paths are never extracted). “Parse without AI” doesn’t extract inline images, so a .md file there always bills exactly 1 page.

---

Indexing

v2

Action	Credits
Exported page	2 per page
Retrieval query	1 per query
Chat turn	100 per turn

These charges are on top of parsing and document storage costs.

v1

Mode	Credits per Page (or Sheet)
Standard	1
Spreadsheet	2
Multi-modal	2

---

Storage

Retained files — files stored without an expiration — are billed daily on each project’s total stored size:

Item	Credits
Retained file storage	100 per GB per day (0.1 per MB)

LlamaCloud snapshots each project’s retained storage once a day and charges it at the rate above. Ephemeral files — files uploaded with an expiration or retention period, including files in ephemeral directories — are never charged for storage.

Retained storage is also capped per project by plan tier; see the plan comparison for the limits and what happens when a project reaches them. To stop paying for a file, delete it, or upload files ephemerally when you don’t need them retained.

---

Extraction

v2 (Tiers)

In the v2 API, total credits per page are the extract tier plus the parse tier.

The default price includes both steps:

Extract tier	Extract credits/page	Default parse tier	Parse credits/page	Default total
Agentic Plus	50	Agentic	10	60
Agentic	15	Agentic	10	25
Cost-effective	5	Cost-effective	3	8

You can choose a different parse tier under Advanced Options. Its cost is added to the extract tier:

Parse tier	Credits/page
Agentic Plus	45
Agentic	10
Cost-effective	3
Fast	1

For text files or files already parsed by LlamaParse, only the extract tier applies. Text files (md, txt, csv, html) use 600 tokens as one page.

Turbo: Turbo prepares documents itself instead of running a separate Parse step, so it has no parse tier and one all-in rate:

Extract tier	Credits/page
Turbo	35

Because Turbo skips parsing, it produces no parse output and accepts no parse tier or parse configuration — those settings are ignored if sent. It does not produce granular bounding boxes. See Tiers for feature availability.

Large schemas: Cost-effective, Agentic, and Agentic Plus support schemas up to 3,200 fields. The extraction rate increases above 250 fields:

Schema size (fields)	Multiplier	Cost-effective credits/page	Agentic credits/page	Agentic Plus credits/page
0–250	×1	5	15	50
251–1,000	×2	10	30	100
1,001–2,000	×3	15	45	150
2,001–3,000	×4	20	60	200
3,001–3,200	×5	25	75	250

The multiplier applies only to the extract tier. Any separate Parse charge still uses the actual page count. Cost-effective and Agentic require extract version 2026-09-14 or later for large schemas.

Schema size is the number of leaf fields in the schema you submit, not a count of what the document returns. A table with 40 rows and 10 columns is 10 fields. See schema size for how nesting, arrays, and $ref are counted.

Spreadsheet mode: Spreadsheet mode reads an Excel or CSV workbook directly. It is available on Cost-effective, Agentic, and Agentic Plus. Each tier uses its extract rate with no separate Parse charge:

Extract tier	Credits per billed page
Cost-effective	5
Agentic	15
Agentic Plus	50

A CSV is billed as a workbook rather than under the text-file rule above. A workbook has no fixed pages, so the billed page count is 10 + ceil(non_empty_cells / 1000), independent of the number of worksheets. A workbook with 1–1,000 non-empty cells bills 11 pages: 55 credits on Cost-effective, 165 on Agentic, or 550 on Agentic Plus. Read metadata.usage.num_pages_billed from the completed job for the exact count charged.

Cost-effective and Agentic require extract version 2026-09-14 or later for spreadsheet mode.

v1 (Modes)

Mode	Credits per Page	Credits per Page (extract only)*
Fast	5	4
Balanced	10	7
Multimodal	20	14
Premium	60	15

*For text files or pre-cached parsed files.

• For text files (md, txt, csv, html), a page is defined as the equivalent of 600 tokens.
• For text files or if the file has previously been parsed by LlamaParse, only extraction costs apply.
• For Multimodal mode, 6 additional credits will be charged for docx/pptx format.

---

Split

Mode	Credits per Page
Default	4 (3 for cached files)

If the file is already present in the LlamaParse cache, only split costs apply.

---

Classification

Mode	Credits per Page
Fast	1
Multimodal	2

---

Cost Optimization Strategies

Parse

1. Use caching — Parsed files are cached for 48 hours. Re-parsing the same file within that window is free.
2. Choose the right tier — Start with Cost-effective (3 credits) for initial testing. Only move to Agentic (10) or Agentic Plus (45) when document complexity requires it. Fast (1 credit) outputs spatial text only — no markdown.
3. Use page ranges — Parse only the pages you need instead of entire documents. Specify target_pages to avoid processing irrelevant content.

Cross-product

4. Leverage extract-only pricing — If a file was previously parsed, subsequent extractions cost less (extract-only rates apply). Parse once, extract many times.
5. Pre-filter with Classify — Use Classify (1-2 credits/page) to filter documents before running more expensive Parse or Extract jobs.
6. Batch strategically — Group similar documents in a single job when possible rather than submitting many small requests.