ModelVerdict

Comparison

Claude Sonnet 5.5 vs Qwen3.8 Flash

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

Verdict

Claude Sonnet 5.5 is more accurate; Qwen3.8 Flash is cheaper in 30 of 30 combinations.

More accurate

Claude Sonnet 5.5

wins 29 of 30

Cheaper

Qwen3.8 Flash

cheaper in 30 of 30

Faster

Claude Sonnet 5.5

2.1 sec vs 14.1 sec

Fewer fixes

Claude Sonnet 5.5

7 vs 67 documents per 1,000

Across 30 task and document-language combinations (speed and fixes: median).

At a glance

English (US)
Accuracy (correct fields)Per 1,000 documents
Task · documentsClaude Sonnet 5.5Qwen3.8 FlashDifferenceClaude Sonnet 5.5Qwen3.8 FlashDifference
Invoice extraction100.00%82.36%▼ 17.64$9.54$1.15−88 %
Email triage98.89%97.48%▼ 1.41$3.21$0.30−91 %
Personal data detection100.00%99.24%▼ 0.76$3.22$0.28−91 %
Contract clauses99.79%95.13%▼ 4.66$4.23$0.62−85 %
Document summaries100.00%99.89%▼ 0.11$3.60$0.75−79 %
Questions about a document100.00%99.78%▼ 0.22$3.84$0.44−89 %
Czech
Accuracy (correct fields)Per 1,000 documents
Task · documentsClaude Sonnet 5.5Qwen3.8 FlashDifferenceClaude Sonnet 5.5Qwen3.8 FlashDifference
Invoice extraction99.95%91.79%▼ 8.16$10.79$1.02−91 %
Email triage98.67%98.52%▼ 0.15$3.92$0.45−89 %
Personal data detection99.71%96.95%▼ 2.76$3.52$0.41−88 %
Contract clauses100.00%99.90%▼ 0.1$5.35$0.63−88 %
Document summaries98.67%95.44%▼ 3.23$5.01$1.48−70 %
Questions about a document100.00%94.22%▼ 5.78$5.03$0.71−86 %
German
Accuracy (correct fields)Per 1,000 documents
Task · documentsClaude Sonnet 5.5Qwen3.8 FlashDifferenceClaude Sonnet 5.5Qwen3.8 FlashDifference
Invoice extraction99.95%93.49%▼ 6.46$11.21$1.12−90 %
Email triage98.00%97.85%▼ 0.15$4.21$0.39−91 %
Personal data detection100.00%99.62%▼ 0.38$3.87$0.36−91 %
Contract clauses100.00%98.00%▼ 2$6.05$0.78−87 %
Document summaries99.11%95.33%▼ 3.78$5.82$1.43−75 %
Questions about a document100.00%98.89%▼ 1.11$5.49$0.46−92 %
Spanish
Accuracy (correct fields)Per 1,000 documents
Task · documentsClaude Sonnet 5.5Qwen3.8 FlashDifferenceClaude Sonnet 5.5Qwen3.8 FlashDifference
Invoice extraction100.00%88.87%▼ 11.13$10.99$1.38−87 %
Email triage97.70%96.74%▼ 0.96$3.76$0.46−88 %
Personal data detection100.00%99.81%▼ 0.19$3.78$0.38−90 %
Contract clauses100.00%100.00%±0$5.70$0.60−90 %
Document summaries95.56%94.44%▼ 1.12$5.55$1.22−78 %
Questions about a document100.00%99.67%▼ 0.33$4.89$0.53−89 %
French
Accuracy (correct fields)Per 1,000 documents
Task · documentsClaude Sonnet 5.5Qwen3.8 FlashDifferenceClaude Sonnet 5.5Qwen3.8 FlashDifference
Invoice extraction99.95%90.46%▼ 9.49$10.71$1.28−88 %
Email triage97.78%96.44%▼ 1.34$3.85$0.45−88 %
Personal data detection100.00%99.71%▼ 0.29$3.81$0.40−90 %
Contract clauses99.95%99.49%▼ 0.46$5.71$0.66−88 %
Document summaries97.89%96.33%▼ 1.56$4.89$1.46−70 %
Questions about a document100.00%98.78%▼ 1.22$4.90$0.57−88 %

Difference: Qwen3.8 Flash vs. Claude Sonnet 5.5; accuracy in percentage points, price in percent. The better value is in bold. Click a task to see the details.

Invoice extraction · English (US) (150 documents)

Claude Sonnet 5.5 is more accurate by 17.6 pp. Qwen3.8 Flash is 8.3× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields100.00%82.36%
Error-free documents100.00%57.33%
Per 1,000 documents$9.54$1.15
To fix / 1,0000427
Speed (median)3.5 sec17.4 sec
Valid answers100.00%90.67%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Invoice extraction · Czech (150 documents)

Claude Sonnet 5.5 is more accurate by 8.2 pp. Qwen3.8 Flash is 10.6× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields99.95%91.79%
Error-free documents99.33%76.67%
Per 1,000 documents$10.79$1.02
To fix / 1,0007233
Speed (median)4.4 sec17.0 sec
Valid answers100.00%96.67%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Invoice extraction · German (150 documents)

Claude Sonnet 5.5 is more accurate by 6.5 pp. Qwen3.8 Flash is 10× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields99.95%93.49%
Error-free documents99.33%87.33%
Per 1,000 documents$11.21$1.12
To fix / 1,0007127
Speed (median)6.0 sec19.0 sec
Valid answers100.00%94.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Invoice extraction · Spanish (150 documents)

Claude Sonnet 5.5 is more accurate by 11.1 pp. Qwen3.8 Flash is 7.9× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields100.00%88.87%
Error-free documents100.00%74.67%
Per 1,000 documents$10.99$1.38
To fix / 1,0000253
Speed (median)5.7 sec22.0 sec
Valid answers100.00%90.67%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Invoice extraction · French (150 documents)

Claude Sonnet 5.5 is more accurate by 9.5 pp. Qwen3.8 Flash is 8.4× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields99.95%90.46%
Error-free documents99.33%83.33%
Per 1,000 documents$10.71$1.28
To fix / 1,0007167
Speed (median)4.9 sec19.4 sec
Valid answers100.00%94.67%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Email triage · English (US) (150 documents)

Claude Sonnet 5.5 is more accurate by 1.4 pp. Qwen3.8 Flash is 10.6× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields98.89%97.48%
Error-free documents90.00%78.67%
Per 1,000 documents$3.21$0.30
To fix / 1,000100213
Speed (median)2.0 sec8.9 sec
Valid answers100.00%100.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Email triage · Czech (150 documents)

Claude Sonnet 5.5 is more accurate by 0.2 pp. Qwen3.8 Flash is 8.8× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields98.67%98.52%
Error-free documents88.00%86.67%
Per 1,000 documents$3.92$0.45
To fix / 1,000120133
Speed (median)2.0 sec13.0 sec
Valid answers100.00%100.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Email triage · German (150 documents)

Claude Sonnet 5.5 is more accurate by 0.1 pp. Qwen3.8 Flash is 10.8× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields98.00%97.85%
Error-free documents83.33%81.33%
Per 1,000 documents$4.21$0.39
To fix / 1,000167187
Speed (median)1.8 sec12.1 sec
Valid answers100.00%100.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Email triage · Spanish (150 documents)

Claude Sonnet 5.5 is more accurate by 1 pp. Qwen3.8 Flash is 8.2× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields97.70%96.74%
Error-free documents80.67%72.00%
Per 1,000 documents$3.76$0.46
To fix / 1,000193280
Speed (median)1.8 sec11.2 sec
Valid answers100.00%100.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Email triage · French (150 documents)

Claude Sonnet 5.5 is more accurate by 1.3 pp. Qwen3.8 Flash is 8.5× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields97.78%96.44%
Error-free documents82.00%76.00%
Per 1,000 documents$3.85$0.45
To fix / 1,000180240
Speed (median)1.7 sec13.2 sec
Valid answers100.00%99.33%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

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

Claude Sonnet 5.5 is more accurate by 0.8 pp. Qwen3.8 Flash is 11.5× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields100.00%99.24%
Error-free documents100.00%98.67%
Per 1,000 documents$3.22$0.28
To fix / 1,000013
Speed (median)1.9 sec7.5 sec
Valid answers100.00%100.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Personal data detection · Czech (150 documents)

Claude Sonnet 5.5 is more accurate by 2.8 pp. Qwen3.8 Flash is 8.7× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields99.71%96.95%
Error-free documents98.00%88.67%
Per 1,000 documents$3.52$0.41
To fix / 1,00020113
Speed (median)2.2 sec9.6 sec
Valid answers100.00%99.33%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Personal data detection · German (150 documents)

Claude Sonnet 5.5 is more accurate by 0.4 pp. Qwen3.8 Flash is 10.8× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields100.00%99.62%
Error-free documents100.00%99.33%
Per 1,000 documents$3.87$0.36
To fix / 1,00007
Speed (median)1.9 sec12.2 sec
Valid answers100.00%100.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Personal data detection · Spanish (150 documents)

Claude Sonnet 5.5 is more accurate by 0.2 pp. Qwen3.8 Flash is 10× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields100.00%99.81%
Error-free documents100.00%98.67%
Per 1,000 documents$3.78$0.38
To fix / 1,000013
Speed (median)1.9 sec10.2 sec
Valid answers100.00%100.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Personal data detection · French (150 documents)

Claude Sonnet 5.5 is more accurate by 0.3 pp. Qwen3.8 Flash is 9.6× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields100.00%99.71%
Error-free documents100.00%98.00%
Per 1,000 documents$3.81$0.40
To fix / 1,000020
Speed (median)2.0 sec10.9 sec
Valid answers100.00%100.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Contract clauses · English (US) (150 documents)

Claude Sonnet 5.5 is more accurate by 4.7 pp. Qwen3.8 Flash is 6.8× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields99.79%95.13%
Error-free documents98.00%69.33%
Per 1,000 documents$4.23$0.62
To fix / 1,00020307
Speed (median)2.1 sec14.1 sec
Valid answers100.00%100.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Contract clauses · Czech (150 documents)

Claude Sonnet 5.5 is more accurate by 0.1 pp. Qwen3.8 Flash is 8.5× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields100.00%99.90%
Error-free documents100.00%98.67%
Per 1,000 documents$5.35$0.63
To fix / 1,000013
Speed (median)2.4 sec13.6 sec
Valid answers100.00%100.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Contract clauses · German (150 documents)

Claude Sonnet 5.5 is more accurate by 2 pp. Qwen3.8 Flash is 7.8× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields100.00%98.00%
Error-free documents100.00%87.33%
Per 1,000 documents$6.05$0.78
To fix / 1,0000127
Speed (median)2.7 sec19.7 sec
Valid answers100.00%99.33%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Contract clauses · Spanish (150 documents)

Both are equally accurate. Qwen3.8 Flash is 9.6× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields100.00%100.00%
Error-free documents100.00%100.00%
Per 1,000 documents$5.70$0.60
To fix / 1,00000
Speed (median)2.7 sec14.2 sec
Valid answers100.00%100.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Contract clauses · French (150 documents)

Claude Sonnet 5.5 is more accurate by 0.5 pp. Qwen3.8 Flash is 8.7× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields99.95%99.49%
Error-free documents99.33%93.33%
Per 1,000 documents$5.71$0.66
To fix / 1,000767
Speed (median)2.6 sec16.2 sec
Valid answers100.00%100.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Document summaries · English (US) (150 documents)

Claude Sonnet 5.5 is more accurate by 0.1 pp. Qwen3.8 Flash is 4.8× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields100.00%99.89%
Error-free documents100.00%99.33%
Per 1,000 documents$3.60$0.75
To fix / 1,00007
Speed (median)1.8 sec17.8 sec
Valid answers100.00%100.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Document summaries · Czech (150 documents)

Claude Sonnet 5.5 is more accurate by 3.2 pp. Qwen3.8 Flash is 3.4× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields98.67%95.44%
Error-free documents92.00%93.33%
Per 1,000 documents$5.01$1.48
To fix / 1,0008067
Speed (median)2.7 sec34.2 sec
Valid answers100.00%96.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Document summaries · German (150 documents)

Claude Sonnet 5.5 is more accurate by 3.8 pp. Qwen3.8 Flash is 4.1× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields99.11%95.33%
Error-free documents94.67%95.33%
Per 1,000 documents$5.82$1.43
To fix / 1,0005347
Speed (median)2.8 sec29.5 sec
Valid answers100.00%95.33%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Document summaries · Spanish (150 documents)

Claude Sonnet 5.5 is more accurate by 1.1 pp. Qwen3.8 Flash is 4.5× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields95.56%94.44%
Error-free documents73.33%93.33%
Per 1,000 documents$5.55$1.22
To fix / 1,00026767
Speed (median)2.1 sec23.9 sec
Valid answers100.00%94.67%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Document summaries · French (150 documents)

Claude Sonnet 5.5 is more accurate by 1.6 pp. Qwen3.8 Flash is 3.4× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields97.89%96.33%
Error-free documents87.33%94.67%
Per 1,000 documents$4.89$1.46
To fix / 1,00012753
Speed (median)2.7 sec28.1 sec
Valid answers100.00%96.67%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Questions about a document · English (US) (150 documents)

Claude Sonnet 5.5 is more accurate by 0.2 pp. Qwen3.8 Flash is 8.7× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields100.00%99.78%
Error-free documents100.00%98.67%
Per 1,000 documents$3.84$0.44
To fix / 1,000013
Speed (median)1.5 sec8.7 sec
Valid answers100.00%100.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Questions about a document · Czech (150 documents)

Claude Sonnet 5.5 is more accurate by 5.8 pp. Qwen3.8 Flash is 7.1× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields100.00%94.22%
Error-free documents100.00%84.67%
Per 1,000 documents$5.03$0.71
To fix / 1,0000153
Speed (median)2.1 sec14.3 sec
Valid answers100.00%98.67%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Questions about a document · German (150 documents)

Claude Sonnet 5.5 is more accurate by 1.1 pp. Qwen3.8 Flash is 11.9× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields100.00%98.89%
Error-free documents100.00%96.67%
Per 1,000 documents$5.49$0.46
To fix / 1,000033
Speed (median)1.9 sec8.6 sec
Valid answers100.00%100.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Questions about a document · Spanish (150 documents)

Claude Sonnet 5.5 is more accurate by 0.3 pp. Qwen3.8 Flash is 9.3× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields100.00%99.67%
Error-free documents100.00%98.67%
Per 1,000 documents$4.89$0.53
To fix / 1,000013
Speed (median)1.4 sec10.3 sec
Valid answers100.00%100.00%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Questions about a document · French (150 documents)

Claude Sonnet 5.5 is more accurate by 1.2 pp. Qwen3.8 Flash is 8.6× cheaper.

MetricClaude Sonnet 5.5Qwen3.8 Flash
Correct fields100.00%98.78%
Error-free documents100.00%97.33%
Per 1,000 documents$4.90$0.57
To fix / 1,000027
Speed (median)1.3 sec10.4 sec
Valid answers100.00%99.33%

Free with login

Accuracy field by field and the documents where the models differ.

Just your email, no password.

Log in with email

We use your email only to log you in (and for the digest if you tick it). No spam.