ModelVerdict

Comparison

GPT-6 Luna vs Mistral Medium 3.5

Same documents, same prompt: where each model is right, what it really costs and how fast it answers.

At a glance

Task · documentsEnglish (US)CzechGermanSpanishFrench
Invoice extraction99.7% / 97.8%99.3% / 93.5%99.2% / 97.4%99.1% / 97.0%98.9% / 98.2%
Email triage97.5% / 94.9%97.9% / 94.5%97.0% / 92.7%——
Personal data detection99.6% / 99.7%99.6% / 99.8%99.6% / 100.0%——
Contract clauses99.8% / 98.1%100.0% / 100.0%100.0% / 97.1%——

Correct fields: GPT-6 Luna / Mistral Medium 3.5; the better one in bold. Click a cell to see the details.

Invoice extraction · English (US) (150 documents)

GPT-6 Luna is more accurate by 1.9 pp. GPT-6 Luna is 8.6× cheaper.

MetricGPT-6 LunaMistral Medium 3.5
Correct fields99.7%97.8%
Error-free documents97%79%
Per 1,000 documents$0.51$4.37
To fix / 1,00033207
Speed (median)3.4 sec2.3 sec
Valid answers100.0%100.0%

Accuracy per field

FieldGPT-6 LunaMistral Medium 3.5
Supplier100.0%100.0%
Supplier ID (EIN, IČO…)100.0%100.0%
Invoice number99.3%100.0%
Payment reference98.0%94.7%
Invoice date100.0%98.7%
Due date100.0%99.3%
Net by tax rate100.0%88.0%
Tax by rate100.0%98.7%
Total100.0%100.0%
Currency100.0%100.0%
Bank account99.3%96.0%

Documents where they differ

Invoice extraction · Czech (150 documents)

GPT-6 Luna is more accurate by 5.8 pp. GPT-6 Luna is 9.3× cheaper.

MetricGPT-6 LunaMistral Medium 3.5
Correct fields99.3%93.5%
Error-free documents93%63%
Per 1,000 documents$0.51$4.69
To fix / 1,00067367
Speed (median)3.2 sec2.6 sec
Valid answers100.0%100.0%

Accuracy per field

FieldGPT-6 LunaMistral Medium 3.5
Supplier95.3%90.0%
Supplier ID (EIN, IČO…)98.7%87.3%
VAT ID98.0%85.3%
Invoice number100.0%99.3%
Payment reference100.0%95.3%
Invoice date100.0%88.7%
Tax point100.0%86.7%
Due date99.3%90.7%
Net by tax rate100.0%99.3%
Tax by rate100.0%98.7%
Total100.0%100.0%
Currency100.0%100.0%
Bank account99.3%94.0%

Documents where they differ

Invoice extraction · German (150 documents)

GPT-6 Luna is more accurate by 1.8 pp. GPT-6 Luna is 8.7× cheaper.

MetricGPT-6 LunaMistral Medium 3.5
Correct fields99.2%97.4%
Error-free documents91%75%
Per 1,000 documents$0.54$4.72
To fix / 1,00087253
Speed (median)3.7 sec2.5 sec
Valid answers100.0%100.0%

Accuracy per field

FieldGPT-6 LunaMistral Medium 3.5
Supplier100.0%97.3%
Supplier ID (EIN, IČO…)98.7%98.7%
VAT ID99.3%96.0%
Invoice number99.3%100.0%
Payment reference100.0%92.7%
Invoice date98.7%98.0%
Tax point100.0%90.7%
Due date100.0%97.3%
Net by tax rate100.0%98.0%
Tax by rate100.0%100.0%
Total100.0%100.0%
Currency100.0%100.0%
Bank account94.0%97.3%

Documents where they differ

Invoice extraction · Spanish (100 documents)

GPT-6 Luna is more accurate by 2.1 pp. GPT-6 Luna is 8.6× cheaper.

MetricGPT-6 LunaMistral Medium 3.5
Correct fields99.1%97.0%
Error-free documents88%67%
Per 1,000 documents$0.55$4.75
To fix / 1,000120330
Speed (median)3.9 sec2.5 sec
Valid answers100.0%100.0%

Accuracy per field

FieldGPT-6 LunaMistral Medium 3.5
Supplier99.0%99.0%
Supplier ID (EIN, IČO…)100.0%100.0%
VAT ID100.0%94.0%
Invoice number100.0%100.0%
Payment reference98.0%93.0%
Invoice date100.0%100.0%
Tax point100.0%93.0%
Due date100.0%97.0%
Net by tax rate100.0%85.0%
Tax by rate100.0%100.0%
Total100.0%100.0%
Currency100.0%100.0%
Bank account91.0%100.0%

Documents where they differ

Invoice extraction · French (98 documents)

GPT-6 Luna is more accurate by 0.7 pp. GPT-6 Luna is 8.5× cheaper.

MetricGPT-6 LunaMistral Medium 3.5
Correct fields98.9%98.2%
Error-free documents86%81%
Per 1,000 documents$0.57$4.81
To fix / 1,000143190
Speed (median)3.8 sec2.6 sec
Valid answers100.0%100.0%

Accuracy per field

FieldGPT-6 LunaMistral Medium 3.5
Supplier99.0%99.0%
Supplier ID (EIN, IČO…)99.0%99.0%
VAT ID98.0%98.0%
Invoice number99.0%100.0%
Payment reference100.0%94.0%
Invoice date100.0%100.0%
Tax point100.0%100.0%
Due date100.0%100.0%
Net by tax rate100.0%95.0%
Tax by rate100.0%98.0%
Total100.0%99.0%
Currency100.0%96.0%
Bank account90.8%98.0%

Documents where they differ

Email triage · English (US) (150 documents)

GPT-6 Luna is more accurate by 2.6 pp. GPT-6 Luna is 10.9× cheaper.

MetricGPT-6 LunaMistral Medium 3.5
Correct fields97.5%94.9%
Error-free documents77%57%
Per 1,000 documents$0.13$1.47
To fix / 1,000227427
Speed (median)2.5 sec0.8 sec
Valid answers100.0%100.0%

Accuracy per field

FieldGPT-6 LunaMistral Medium 3.5
Category100.0%100.0%
Priority98.0%63.3%
Sentiment79.3%90.7%
Customer name100.0%100.0%
Order number100.0%100.0%
Invoice number100.0%100.0%
Amount100.0%100.0%
Currency100.0%100.0%
Deadline100.0%100.0%

Documents where they differ

  • Danielle Little · Email · GPT-6 Luna 9/9, Mistral Medium 3.5 7/9
  • Omar Cabrera · Email · GPT-6 Luna 8/9, Mistral Medium 3.5 9/9
  • Russell Bates · Email · GPT-6 Luna 9/9, Mistral Medium 3.5 8/9
  • Kyle Schwartz · Email · GPT-6 Luna 9/9, Mistral Medium 3.5 8/9
  • James Arnold · Email · GPT-6 Luna 9/9, Mistral Medium 3.5 8/9
  • Gene Everett · Email · GPT-6 Luna 9/9, Mistral Medium 3.5 8/9
  • Cody Baldwin · Email · GPT-6 Luna 8/9, Mistral Medium 3.5 7/9
  • Dale Sanchez · Email · GPT-6 Luna 9/9, Mistral Medium 3.5 8/9

Email triage · Czech (150 documents)

GPT-6 Luna is more accurate by 3.4 pp. GPT-6 Luna is 10.9× cheaper.

MetricGPT-6 LunaMistral Medium 3.5
Correct fields97.9%94.5%
Error-free documents81%55%
Per 1,000 documents$0.18$1.95
To fix / 1,000187447
Speed (median)2.6 sec0.8 sec
Valid answers100.0%100.0%

Accuracy per field

FieldGPT-6 LunaMistral Medium 3.5
Category100.0%100.0%
Priority98.7%63.3%
Sentiment82.7%87.3%
Customer name100.0%100.0%
Order number100.0%100.0%
Invoice number100.0%100.0%
Amount100.0%100.0%
Currency100.0%100.0%
Deadline100.0%100.0%

Documents where they differ

Email triage · German (150 documents)

GPT-6 Luna is more accurate by 4.2 pp. GPT-6 Luna is 11.6× cheaper.

MetricGPT-6 LunaMistral Medium 3.5
Correct fields97.0%92.7%
Error-free documents73%48%
Per 1,000 documents$0.15$1.77
To fix / 1,000273520
Speed (median)2.8 sec0.8 sec
Valid answers100.0%100.0%

Accuracy per field

FieldGPT-6 LunaMistral Medium 3.5
Category100.0%100.0%
Priority99.3%56.7%
Sentiment73.3%78.0%
Customer name100.0%100.0%
Order number100.0%100.0%
Invoice number100.0%100.0%
Amount100.0%100.0%
Currency100.0%100.0%
Deadline100.0%100.0%

Documents where they differ

Personal data detection · English (US) (150 documents)

Mistral Medium 3.5 is more accurate by 0.1 pp. GPT-6 Luna is 10.8× cheaper.

MetricGPT-6 LunaMistral Medium 3.5
Correct fields99.6%99.7%
Error-free documents97%98%
Per 1,000 documents$0.11$1.15
To fix / 1,0002720
Speed (median)1.8 sec0.8 sec
Valid answers100.0%100.0%

Accuracy per field

FieldGPT-6 LunaMistral Medium 3.5
Names98.7%99.3%
Email addresses100.0%100.0%
Phone numbers100.0%100.0%
Addresses98.7%98.7%
Dates of birth100.0%100.0%
National IDs100.0%100.0%
Bank accounts100.0%100.0%

Documents where they differ

Personal data detection · Czech (150 documents)

Mistral Medium 3.5 is more accurate by 0.2 pp. GPT-6 Luna is 11.7× cheaper.

MetricGPT-6 LunaMistral Medium 3.5
Correct fields99.6%99.8%
Error-free documents97%99%
Per 1,000 documents$0.12$1.40
To fix / 1,0002713
Speed (median)1.8 sec0.8 sec
Valid answers100.0%100.0%

Accuracy per field

FieldGPT-6 LunaMistral Medium 3.5
Names100.0%98.7%
Email addresses100.0%100.0%
Phone numbers100.0%100.0%
Addresses97.3%100.0%
Dates of birth100.0%100.0%
National IDs100.0%100.0%
Bank accounts100.0%100.0%

Documents where they differ

Personal data detection · German (150 documents)

Mistral Medium 3.5 is more accurate by 0.4 pp. GPT-6 Luna is 11.8× cheaper.

MetricGPT-6 LunaMistral Medium 3.5
Correct fields99.6%100.0%
Error-free documents97%100%
Per 1,000 documents$0.12$1.38
To fix / 1,000270
Speed (median)2.2 sec0.9 sec
Valid answers100.0%100.0%

Accuracy per field

FieldGPT-6 LunaMistral Medium 3.5
Names100.0%100.0%
Email addresses100.0%100.0%
Phone numbers100.0%100.0%
Addresses97.3%100.0%
Dates of birth100.0%100.0%
National IDs100.0%100.0%
Bank accounts100.0%100.0%

Documents where they differ

Contract clauses · English (US) (150 documents)

GPT-6 Luna is more accurate by 1.7 pp. GPT-6 Luna is 11.7× cheaper.

MetricGPT-6 LunaMistral Medium 3.5
Correct fields99.8%98.1%
Error-free documents99%87%
Per 1,000 documents$0.20$2.35
To fix / 1,00013127
Speed (median)2.7 sec1.1 sec
Valid answers100.0%100.0%

Accuracy per field

FieldGPT-6 LunaMistral Medium 3.5
Provider98.7%87.3%
Provider ID100.0%100.0%
Client98.7%88.7%
Client ID100.0%99.3%
Contract type100.0%100.0%
Price (net)100.0%100.0%
Currency100.0%100.0%
Billing period100.0%100.0%
Effective date100.0%100.0%
End date100.0%100.0%
Notice period (months)100.0%100.0%
Late-payment penalty (% per day)100.0%100.0%
Governing law100.0%100.0%

Documents where they differ

Contract clauses · Czech (150 documents)

Both are equally accurate. GPT-6 Luna is 11.3× cheaper.

MetricGPT-6 LunaMistral Medium 3.5
Correct fields100.0%100.0%
Error-free documents100%100%
Per 1,000 documents$0.27$3.06
To fix / 1,00000
Speed (median)2.6 sec1.2 sec
Valid answers100.0%100.0%

Accuracy per field

FieldGPT-6 LunaMistral Medium 3.5
Provider100.0%100.0%
Provider ID100.0%100.0%
Client100.0%100.0%
Client ID100.0%100.0%
Contract type100.0%100.0%
Price (net)100.0%100.0%
Currency100.0%100.0%
Billing period100.0%100.0%
Effective date100.0%100.0%
End date100.0%100.0%
Notice period (months)100.0%100.0%
Late-payment penalty (% per day)100.0%100.0%
Governing law100.0%100.0%

Documents where they differ

Both models got the same number of fields right on every example document.

Contract clauses · German (150 documents)

GPT-6 Luna is more accurate by 2.9 pp. GPT-6 Luna is 10.4× cheaper.

MetricGPT-6 LunaMistral Medium 3.5
Correct fields100.0%97.1%
Error-free documents100%75%
Per 1,000 documents$0.27$2.76
To fix / 1,0000247
Speed (median)3.0 sec1.2 sec
Valid answers100.0%100.0%

Accuracy per field

FieldGPT-6 LunaMistral Medium 3.5
Provider100.0%100.0%
Provider ID100.0%82.0%
Client100.0%100.0%
Client ID100.0%80.7%
Contract type100.0%100.0%
Price (net)100.0%100.0%
Currency100.0%100.0%
Billing period100.0%100.0%
Effective date100.0%100.0%
End date100.0%100.0%
Notice period (months)100.0%100.0%
Late-payment penalty (% per day)100.0%100.0%
Governing law100.0%100.0%

Documents where they differ