Interfaze vs Mistral OCR: Document-by-Document Results
Dense tiny-print table
Same page and field list for both APIs. Mistral OCR got 8 of 20 fields wrong, interfaze-1 got 0.
Boxes on the page
Returned JSON
correct value
wrong value, boxed where the API read it
no box from the API: boxed where the value is
○
not returned
Interfazeinterfaze-1structured JSON
20/20 field checks passed
✓
pr8406_amount
125,000
0.99
✓
pr8406_payment_number
10,270,279
0.95
✓
ref_600082pt_description
SUPORT PRODUCT THEATER
1.00
✓
ref_cr5483_description
Static Signs f/Meeting Foyer
0.98
✓
ref_cr5483_amount
3,500
0.75
✓
rems_strategies_amount
20,401.22
0.87
✓
rems_strategies_payment_number
100,022,462
0.90
✓
sheet_label
Sheet 1
0.94
✓
bates_number
s2465-0001
0.99
✓
exhibit_stamp_number
s2465
0.99
✓
exhibit_stamp_caption
State of OK v. Purdue
0.97
✓
first_reference
23FPPSE201
0.99
✓
cr5034_description
AfPA Membership
0.97
✓
cr5086_description
CCTA-Membership
0.98
✓
pr5460_description
Sponsorship
0.99
✓
pr6747_description
Corporate Membership
0.99
✓
s010614_description
ASSOCIATE SUSTAINING MEMBERSHIP
0.99
✓
ref_20472_description
AAPHM 2015 Annual Assembly
0.97
✓
ref_58221_description
APINV-COUNCIL MBMRSHP DUES
0.99
✓
pr1864_date
11/16/2015
0.99
Mistralmistral-ocr-lateststructured JSON
12/20 field checks passed
✗
ref_cr5483_description
Static Signs F/Meeting Fever · page says Static Signs f/Meeting Foyer
No box returned. The box marks where the value is on the page.
✗
exhibit_stamp_caption
State of OK • Purdue · page says State of OK v. Purdue
0.74
✗
first_reference
48522 · page says 23FPPSE201
No box returned. The box marks where the value is on the page.
✗
ref_58221_description
AFINV-COUNCIL MBRSHP DUES · page says APINV-COUNCIL MBMRSHP DUES
No box returned. The box marks where the value is on the page.
○
pr8406_amount
not returned · page says 125,000
Not returned. The box marks where the value is on the page.
○
pr8406_payment_number
not returned · page says 10,270,279
Not returned. The box marks where the value is on the page.
○
cr5034_description
not returned · page says AfPA Membership
Not returned. The box marks where the value is on the page.
○
cr5086_description
not returned · page says CCTA-Membership
Not returned. The box marks where the value is on the page.
✓
ref_600082pt_description
SUPORT PRODUCT THEATER
Read correctly.
✓
ref_cr5483_amount
3,500
Read correctly.
✓
rems_strategies_amount
20,401.22
Read correctly.
✓
rems_strategies_payment_number
100,022,462
Read correctly.
✓
sheet_label
Sheet 1
1.00
✓
bates_number
s2465-0001
1.00
✓
exhibit_stamp_number
s2465
Read correctly.
✓
pr5460_description
Sponsorship
Read correctly.
✓
pr6747_description
Corporate Membership
Read correctly.
✓
s010614_description
ASSOCIATE SUSTAINING MEMBERSHIP
Read correctly.
✓
ref_20472_description
AAPHM 2015 Annual Assembly
Read correctly.
✓
pr1864_date
11/16/2015
Read correctly.
LinkedIn printout with exhibit stamp
Same page and field list for both APIs. Mistral OCR got 7 of 20 fields wrong, interfaze-1 got 0.
Boxes on the page
Returned JSON
correct value
wrong value, boxed where the API read it
no box from the API: boxed where the value is
○
not returned
Interfazeinterfaze-1structured JSON
20/20 field checks passed
✓
print_timestamp
11/15/21, 10:42 AM
0.99
✓
page_title
(5) Interests | Sabrina Solis | LinkedIn
0.94
✓
profile_slug
sabrina-solis-25950979
Read correctly.
✓
page_count
1/7
1.00
✓
exhibit_stamp
AND-NY-01164
0.99
✓
bates_number
AND-NY-01164.00001
0.99
✓
scce_followers
58,590 followers
0.99
✓
forbes_followers
17,271,292 followers
0.99
✓
raps_followers
113,236 followers
0.99
✓
cardinal_followers
447,568 followers
0.99
✓
fifth_company
National Association of Boards of Pharmacy
0.99
✓
ultimate_followers
87,904 followers
0.99
✓
au_followers
920 followers
0.99
✓
nabp_followers
18,074 followers
0.99
✓
second_company
Ultimate Software
0.99
✓
fourth_company
American University Law Online
1.00
✓
sixth_company
Regulatory Affairs Professionals Society (RAPS)
0.99
✓
first_company
Society of Corporate Compliance and Ethics (SCCE)
0.99
✓
active_tab
Companies
1.00
✓
headline_title
Sr. Manager, DEA Compliance
0.99
Mistralmistral-ocr-lateststructured JSON
13/20 field checks passed
✗
active_tab
Interests · page says Companies
No box returned. The box marks where the value is on the page.
○
exhibit_stamp
not returned · page says AND-NY-01164
Not returned. The box marks where the value is on the page.
○
fifth_company
not returned · page says National Association of Boards of Pharmacy
Not returned. The box marks where the value is on the page.
○
second_company
not returned · page says Ultimate Software
Not returned. The box marks where the value is on the page.
○
fourth_company
not returned · page says American University Law Online
Not returned. The box marks where the value is on the page.
○
sixth_company
not returned · page says Regulatory Affairs Professionals Society (RAPS)
Not returned. The box marks where the value is on the page.
○
first_company
not returned · page says Society of Corporate Compliance and Ethics (SCCE)
Not returned. The box marks where the value is on the page.
✓
print_timestamp
11/15/21, 10:42 AM
1.00
✓
page_title
(5) Interests | Sabrina Solis | LinkedIn
1.00
✓
profile_slug
sabrina-solis-25950979
1.00
✓
page_count
1/7
1.00
✓
bates_number
AND-NY-01164.00001
1.00
✓
scce_followers
58,590 followers
1.00
✓
forbes_followers
17,271,292 followers
1.00
✓
raps_followers
113,236 followers
1.00
✓
cardinal_followers
447,568 followers
1.00
✓
ultimate_followers
87,904 followers
1.00
✓
au_followers
920 followers
1.00
✓
nabp_followers
18,074 followers
1.00
✓
headline_title
Sr. Manager, DEA Compliance | Compliance - DEA/Regulatory | Senior Leaders...
1.00
Dark noisy cover scan
Same page and field list for both APIs. Mistral OCR got 4 of 4 fields wrong, interfaze-1 got 0.
Boxes on the page
Returned JSON
correct value
wrong value, boxed where the API read it
no box from the API: boxed where the value is
○
not returned
Interfazeinterfaze-1structured JSON
4/4 field checks passed
✓
label_title
L I S T I N G
0.66
✓
label_code
( H - 4 )
0.97
✓
handwritten_id
MRS 21155
0.96
✓
stamp_text
CONSUMER RESEARCH LIBRARY
Read correctly.
Mistralmistral-ocr-lateststructured JSON
0/4 field checks passed
✗
label_title
H - 4 · page says LISTING
0.91
✗
label_code
MRS 21155 · page says (H-4)
0.94
○
handwritten_id
not returned · page says MRS 21155
Not returned. The box marks where the value is on the page.
○
stamp_text
not returned · page says CONSUMER RESEARCH LIBRARY
Not returned.
Newspaper op-ed, multi-column
Same page and field list for both APIs. Mistral OCR got 3 of 22 fields wrong, interfaze-1 got 0.
Boxes on the page
Returned JSON
correct value
wrong value, boxed where the API read it
no box from the API: boxed where the value is
○
not returned
Interfazeinterfaze-1structured JSON
22/22 field checks passed
✓
headline
Climate of Fear
1.00
✓
byline
By Richard Lindzen
0.99
✓
publication_name
The Wall Street Journal
0.99
✓
section_label
OPINION
0.99
✓
issue_date
WEDNESDAY, APRIL 12, 2006
1.00
✓
author_affiliation
Mr. Lindzen is Alfred P. Sloan Professor of Atmospheric Science at MIT.
0.97
✓
climate_research_spending_today
$1.7 billion
0.99
✓
co2_increase_percent
30%
0.96
✓
warmest_year_claimed
1998
0.99
✓
hearings_year
1992
0.99
✓
named_effect
Iris Effect
0.93
✓
photo_credit
Images.com/Corbis
0.59
✓
tv_host
Ted Koppel
0.99
✓
libel_author
Ross Gelbspan
0.99
✓
dismissed_director
Henk Tennekes
0.99
✓
wmo_director
Aksel Winn-Nielsen
Read correctly.
✓
ipcc_first_head
Bert Bolin
0.97
✓
debate_year
1991
0.99
✓
ipcc_claim_source
Sir John Houghton
0.99
✓
texas_rep
Joe Barton
0.99
✓
paleoclimatologist
Michael Mann
0.99
✓
climate_plan_year
2003
0.99
Mistralmistral-ocr-lateststructured JSON
19/22 field checks passed
✗
warmest_year_claimed
1988 · page says 1998
No box returned. The box marks where the value is on the page.
✗
libel_author
Ross Geibspan · page says Ross Gelbspan
No box returned. The box marks where the value is on the page.
○
photo_credit
not returned · page says Images.com/Corbis
Not returned. The box marks where the value is on the page.
✓
headline
Climate of Fear
0.99
✓
byline
By Richard Lindzen
1.00
✓
publication_name
The Wall Street Journal
1.00
✓
section_label
OPINION
1.00
✓
issue_date
WEDNESDAY, APRIL 12, 2006
1.00
✓
author_affiliation
Mr. Lindzen is Alfred P. Sloan Professor of Atmospheric Science at MIT.
1.00
✓
climate_research_spending_today
$1.7 billion
Read correctly.
✓
co2_increase_percent
about 30%
Read correctly.
✓
hearings_year
1992
Read correctly.
✓
named_effect
"Iris Effect"
Read correctly.
✓
tv_host
Ted Koppel
Read correctly.
✓
dismissed_director
Henk Tennekes
Read correctly.
✓
wmo_director
Aksel Winn-Nielsen
Read correctly.
✓
ipcc_first_head
Bert Bolin
Read correctly.
✓
debate_year
1991
Read correctly.
✓
ipcc_claim_source
Sir John Houghton
Read correctly.
✓
texas_rep
Joe Barton
Read correctly.
✓
paleoclimatologist
Michael Mann
Read correctly.
✓
climate_plan_year
2003
Read correctly.
Scanned slide with outlined hollow text
Same page and field list for both APIs. Mistral OCR got 3 of 13 fields wrong, interfaze-1 got 0.
Boxes on the page
Returned JSON
correct value
wrong value, boxed where the API read it
no box from the API: boxed where the value is
○
not returned
Interfazeinterfaze-1structured JSON
13/13 field checks passed
✓
slide_title
New DPN Pain Indication Represents a Clinical Milestone in Pain Management
1.00
✓
first_box
FIRST and ONLY opioid with DPN Pain indication
Read correctly.
✓
emphasized_word
EFFICACY
0.47
✓
bottom_identifier
s2376-0004
0.99
✓
control_number
JAN-MS-00027224
0.99
✓
brand_team
NUCYNTA ER & NUCYNTA Brand Team
0.98
✓
marketplace
Pain Marketplace
0.99
✓
agency
FDA
1.00
✓
abbreviation
DPN
1.00
✓
slide_footer
For internal training use only. Do not duplicate, modify, distribute, or use this item in a selling situation.
0.82
✓
page_number
4
0.96
✓
confidential_label
Confidential
0.99
✓
pain_types
acute and chronic
0.96
Mistralmistral-ocr-lateststructured JSON
10/13 field checks passed
✗
bottom_identifier
**NUCYNTA ER & NEUROPATHY** · page says s2376-0004
0.71
✗
control_number
s2376-0004 · page says JAN-MS-00027224
1.00
○
agency
not returned · page says FDA
Not returned. The box marks where the value is on the page.
✓
slide_title
**New DPN Pain Indication Represents a Clinical Milestone in Pain Management**
1.00
✓
first_box
FIRST and ONLY opioid with DPN Pain indication
0.96
✓
emphasized_word
EFFICACY
Read correctly.
✓
brand_team
NUCYNTA ER & NUCYNTA Brand Team
Read correctly.
✓
marketplace
Pain Marketplace
Read correctly.
✓
abbreviation
DPN
Read correctly.
✓
slide_footer
For internal training use only. Do not duplicate, modify, distribute, or use this item in a selling situation
0.99
✓
page_number
4
Read correctly.
✓
confidential_label
Confidential
1.00
✓
pain_types
acute and chronic
Read correctly.
Annual report letter page, rotated 270 degrees
Same page and field list for both APIs. Mistral OCR got 3 of 18 fields wrong, interfaze-1 got 0.
Boxes on the page
Returned JSON
correct value
wrong value, boxed where the API read it
no box from the API: boxed where the value is
○
not returned
Interfazeinterfaze-1structured JSON
18/18 field checks passed
✓
heading
Dear Shareholder
0.99
✓
sabmiller_stake
28.6%
0.99
✓
dividend_rate
$3.44 per share
1.00
✓
five_year_return
142.6%
0.96
✓
kraft_ratio
0.7
0.99
✓
bates_number
8009533391
0.99
✓
dividend_increase
7.5%
0.96
✓
total_return_2006
19.9%
0.96
✓
sp500_five_year
35.0%
0.99
✓
record_date
March 16, 2007
0.99
✓
distribution_date
March 30, 2007
0.99
✓
net_revenues
$67 billion
0.99
✓
eps
$4.05
0.98
✓
total_assets
$48 billion
0.99
✓
shareholders_equity
$12.7 billion
0.97
✓
lakson_country
Pakistan
0.99
✓
section_heading
Kraft Spin-Off
0.98
✓
page_number
4
1.00
Mistralmistral-ocr-lateststructured JSON
15/18 field checks passed
✗
dividend_increase
75% · page says 7.5%
No box returned. The box marks where the value is on the page.
✗
total_assets
$4.8 billion · page says $48 billion
No box returned. The box marks where the value is on the page.
○
section_heading
not returned · page says Kraft Spin-Off
Not returned. The box marks where the value is on the page.
✓
heading
Dear Shareholder
1.00
✓
sabmiller_stake
28.6%
Read correctly.
✓
dividend_rate
$3.44 per share
Read correctly.
✓
five_year_return
142.6%
Read correctly.
✓
kraft_ratio
0.7
Read correctly.
✓
bates_number
8009533391
Read correctly.
✓
total_return_2006
19.9%
Read correctly.
✓
sp500_five_year
35.0%
Read correctly.
✓
record_date
March 16, 2007
Read correctly.
✓
distribution_date
March 30, 2007
Read correctly.
✓
net_revenues
$67 billion
Read correctly.
✓
eps
$4.05
Read correctly.
✓
shareholders_equity
$12.7 billion
Read correctly.
✓
lakson_country
Pakistan
Read correctly.
✓
page_number
4
Read correctly.
Upside-down photo of a sign
Same page and field list for both APIs. Mistral OCR got 2 of 5 fields wrong, interfaze-1 got 0.
Boxes on the page
Returned JSON
correct value
wrong value, boxed where the API read it
no box from the API: boxed where the value is
○
not returned
Interfazeinterfaze-1structured JSON
5/5 field checks passed
✓
line_1
D'PUNCAK
0.99
✓
line_2
BERIS
0.99
✓
line_3
LAKE RESORT
0.99
✓
line_4
RESTORAN
0.99
✓
line_4_en
RESTAURANT
Read correctly.
Mistralmistral-ocr-lateststructured JSON
3/5 field checks passed
✗
line_1
PUNCA · page says D'PUNCAK
No box returned. The box marks where the value is on the page.
○
line_4_en
not returned · page says Restaurant
Not returned.
✓
line_2
BERIS
Read correctly.
✓
line_3
LAKE RESORT
Read correctly.
✓
line_4
RESTORAN
Read correctly.
ID card (phone photo)
Same page and field list for both APIs. Mistral OCR got 2 of 13 fields wrong, interfaze-1 got 0.
Boxes on the page
Returned JSON
correct value
wrong value, boxed where the API read it
no box from the API: boxed where the value is
○
not returned
Interfazeinterfaze-1structured JSON
13/13 field checks passed
✓
last_name
BAILEY
0.99
✓
first_name
WESTON COLE
0.99
✓
date_of_birth
05/01/1965
0.98
✓
address
4050 VALLEY PIKE WINCHESTER, VA 22602-2457
0.98
✓
customer_number
T16700588
0.99
✓
expires
05/01/2031
0.99
✓
issued
05/21/2023
0.99
✓
hazmat_expires
05/21/2028
0.99
✓
height
6'-02"
Read correctly.
✓
eye_color
BLU
1.00
✓
license_class
A
0.99
✓
restrictions
B
0.99
✓
document_discriminator
071545852
0.96
Mistralmistral-ocr-lateststructured JSON
11/13 field checks passed
✗
document_discriminator
071845852 · page says 071545852
0.92
○
eye_color
not returned · page says BLU
Not returned. The box marks where the value is on the page.
✓
last_name
BAILEY
1.00
✓
first_name
WESTON COLE
1.00
✓
date_of_birth
05/01/1965
0.99
✓
address
4050 VALLEY PIKE, WINCHESTER, VA 22602-2457
Read correctly.
✓
customer_number
T16700588
1.00
✓
expires
05/01/2031
1.00
✓
issued
05/21/2023
1.00
✓
hazmat_expires
05/21/2028
1.00
✓
height
6'-02"
0.99
✓
license_class
A
Read correctly.
✓
restrictions
B
Read correctly.
Typewritten financial statement
Same page and field list for both APIs. Mistral OCR got 2 of 9 fields wrong, interfaze-1 got 0.
Boxes on the page
Returned JSON
correct value
wrong value, boxed where the API read it
no box from the API: boxed where the value is
○
not returned
Interfazeinterfaze-1structured JSON
9/9 field checks passed
✓
organization
The Tobacco Institute
0.99
✓
membership_dues
$15,200,104
0.98
✓
total_receipts
$15,698,217
0.98
✓
staff_compensation
$ 3,716,658
0.91
✓
legal_fees
1,927,714
0.99
✓
computer_services
589,265
0.99
✓
total_disbursements
$15,497,380
0.79
✓
excess_of_receipts
$ 200,837
0.98
✓
bates_number
TI48331278
0.81
Mistralmistral-ocr-lateststructured JSON
7/9 field checks passed
✗
legal_fees
$1,927,714 · page says 1,927,714
No box returned. The box marks where the value is on the page.
✗
computer_services
$ 589,265 · page says 589,265
No box returned. The box marks where the value is on the page.
✓
organization
The Tobacco Institute
1.00
✓
membership_dues
$15,200,104
Read correctly.
✓
total_receipts
$15,698,217
Read correctly.
✓
staff_compensation
$ 3,716,658
Read correctly.
✓
total_disbursements
$15,497,380
1.00
✓
excess_of_receipts
$ 200,837
1.00
✓
bates_number
TI48331278
1.00
IRS Form 990-PF with stamps
Same page and field list for both APIs. Mistral OCR got 2 of 22 fields wrong, interfaze-1 got 0.
Boxes on the page
Returned JSON
correct value
wrong value, boxed where the API read it
no box from the API: boxed where the value is
○
not returned
Interfazeinterfaze-1structured JSON
22/22 field checks passed
✓
foundation_name
ExxonMobil Foundation
0.99
✓
ein
13-6082357
0.99
✓
room_suite
2441
0.96
✓
telephone
713-656-4048
0.99
✓
fair_market_value
79,260,859
0.88
✓
contributions_received
38,815,770
0.98
✓
gross_sales_price
67,300,000
0.99
✓
total_line_12_a
35,871,538
0.99
✓
grants_paid_d
62,495,330
0.96
✓
net_investment_income
517,640
0.92
✓
received_stamp_date
APR 3 0 2007
0.99
✓
street_address
800 Bell Street
0.98
✓
city_state_zip
Houston, TX 77002
0.94
✓
omb_number
1545-0052
0.99
✓
net_gain_6a
-3,493,622
0.84
✓
total_expenses_24_a
367,605
0.87
✓
contributions_paid_25_a
65,271,530
0.97
✓
total_expenses_26_a
65,639,135
0.94
✓
total_disbursements_26_d
62,831,185
0.90
✓
excess_27a
-29,767,597
0.92
✓
scanned_stamp_date
MAY 03 2007
0.97
✓
received_stamp_city
OGDEN, UT
0.82
Mistralmistral-ocr-lateststructured JSON
20/22 field checks passed
✗
received_stamp_date
MAY 03 2007 · page says APR 30 2007
No box returned. The box marks where the value is on the page.
✗
received_stamp_city
MAY 03 2007 · page says OGDEN, UT
No box returned. The box marks where the value is on the page.
✓
foundation_name
ExxonMobil Foundation
Read correctly.
✓
ein
13-6082357
Read correctly.
✓
room_suite
2441
Read correctly.
✓
telephone
713-656-4048
Read correctly.
✓
fair_market_value
$ 79,260,859
Read correctly.
✓
contributions_received
38,815,770
Read correctly.
✓
gross_sales_price
67,300,000
Read correctly.
✓
total_line_12_a
35,871,538
Read correctly.
✓
grants_paid_d
62,495,330
Read correctly.
✓
net_investment_income
517,640
Read correctly.
✓
street_address
800 Bell Street
Read correctly.
✓
city_state_zip
Houston, TX 77002
Read correctly.
✓
omb_number
1545-0052
0.97
✓
net_gain_6a
-3,493,622
Read correctly.
✓
total_expenses_24_a
367,605
Read correctly.
✓
contributions_paid_25_a
65,271,530
Read correctly.
✓
total_expenses_26_a
65,639,135
Read correctly.
✓
total_disbursements_26_d
62,831,185
Read correctly.
✓
excess_27a
-29,767,597
Read correctly.
✓
scanned_stamp_date
MAY 03 2007
Read correctly.
Dense journal page with sideways download stamp
Same page and field list for both APIs. Mistral OCR got 2 of 18 fields wrong, interfaze-1 got 0.
Boxes on the page
Returned JSON
correct value
wrong value, boxed where the API read it
no box from the API: boxed where the value is
○
not returned
Interfazeinterfaze-1structured JSON
18/18 field checks passed
✓
continued_from
p. 33
Read correctly.
✓
po_box
899
0.99
✓
rugg_page
818
1.00
✓
iso_location
Akron, Ohio
0.98
✓
page_number
44
Read correctly.
✓
download_ip
70.88.140.126
0.92
✓
download_date
09 May 2016
0.99
✓
left_title
SIMPLIFIED METHOD IN FINDING OZONE
Read correctly.
✓
right_title
SOAP-MAKING ODOR CONTROL
0.99
✓
fig_caption
Control of odors from centrifuge room.
0.98
✓
diagram_label_tank
Acid Tank
0.98
✓
diagram_label_supply
Constant Acid Supply
Read correctly.
✓
seminar_theme
The Concept of Total Dose Assessment.
Read correctly.
✓
seminar_contact
Professor Emil T. Chanlett
0.99
✓
ack_2
Richard J. Collins
0.99
✓
bottom_section_title
AIR POLLUTION CONTROL PROGRAM
0.99
✓
journal_footer
Journal of the Air Pollution Control Association
0.99
✓
powell_year
1946
0.99
Mistralmistral-ocr-lateststructured JSON
16/18 field checks passed
✗
rugg_page
24, 818 (May 1952) · page says 818
No box returned. The box marks where the value is on the page.
✗
download_date
11:17 09 May 2016 · page says 09 May 2016
No box returned. The box marks where the value is on the page.
✓
continued_from
(Continued from p. 33)
0.98
✓
po_box
P. O. Box 899
Read correctly.
✓
iso_location
Akron, Ohio
Read correctly.
✓
page_number
44
1.00
✓
download_ip
[70.88.140.126]
Read correctly.
✓
left_title
SIMPLIFIED METHOD IN FINDING OZONE
0.90
✓
right_title
SOAP-MAKING ODOR CONTROL
1.00
✓
fig_caption
Control of odors from centrifuge room.
0.99
✓
diagram_label_tank
Acid Tank
Read correctly.
✓
diagram_label_supply
Constant Acid Supply
Read correctly.
✓
seminar_theme
The Concept of Total Dose Assessment
Read correctly.
✓
seminar_contact
Professor Emil T. Chanlett
Read correctly.
✓
ack_2
Richard J. Collins
Read correctly.
✓
bottom_section_title
AIR POLLUTION CONTROL PROGRAM
1.00
✓
journal_footer
Journal of the Air Pollution Control Association
1.00
✓
powell_year
1946
Read correctly.
Near-blank page with upside-down legal banner
Same page and field list for both APIs. Mistral OCR got 2 of 8 fields wrong, interfaze-1 got 0.
Boxes on the page
Returned JSON
correct value
wrong value, boxed where the API read it
no box from the API: boxed where the value is
○
not returned
Interfazeinterfaze-1structured JSON
8/8 field checks passed
✓
box_label
REFERENCES CODES BY MAIN HEADING
0.99
✓
first_action
500-06-000076-980
0.99
✓
second_action
500-06-000070-983
0.96
✓
banner_company
ITCAN
0.99
✓
side_number
MRS 21155
1.00
✓
banner_province
QUEBEC
0.96
✓
banner_ending
STRICTLY CONFIDENTIAL
0.99
✓
banner_doc_word
DOCUMENT
0.99
Mistralmistral-ocr-lateststructured JSON
6/8 field checks passed
✗
box_label
REFERENCES CODES · page says REFERENCES CODES BY MAIN HEADING
1.00
✗
side_number
983 · page says MRS 21155
No box returned. The box marks where the value is on the page.
✓
first_action
500-06-000076-980
Read correctly.
✓
second_action
500-06-000070-983
Read correctly.
✓
banner_company
ITCAN
Read correctly.
✓
banner_province
QUEBEC
Read correctly.
✓
banner_ending
STRICTLY CONFIDENTIAL
Read correctly.
✓
banner_doc_word
DOCUMENT
Read correctly.
Report text page, rotated 90 degrees
Same page and field list for both APIs. Mistral OCR got 2 of 4 fields wrong, interfaze-1 got 0.
Boxes on the page
Returned JSON
correct value
wrong value, boxed where the API read it
no box from the API: boxed where the value is
○
not returned
Interfazeinterfaze-1structured JSON
4/4 field checks passed
✓
page_number
5
0.99
✓
days
30
1.00
✓
feedback_from
BIRD
0.99
✓
bates_number
55265 2347
0.99
Mistralmistral-ocr-lateststructured JSON
2/4 field checks passed
✗
bates_number
DPO 476741478271 · page says 55265 2347
1.00
○
days
not returned · page says 30
Not returned. The box marks where the value is on the page.
✓
page_number
5
Read correctly.
✓
feedback_from
BIRD
Read correctly.
Send us your hardest PDF.
If we can't beat your current provider,
we'll give you $3,000 in credits.
Start for free. No credit card required.