FinCon Bench

Rank #51 of 54

OpenAI

GPT-OSS 120B

via AWS Bedrock + Ollama Cloud

gpt-oss-120b

Failure rate

36%

197 of 550 decided probes failed.

Coverage

100%

Share of probes the judges actually decided.

Spread across passes

5.3 pts

Highest minus lowest pass rate over 173 repeated probes.

Compliance / behaviour

60% / 65%

Cost per pass

621 avg reply tokens, —s.

Failure rate by category

Lower is better. The count beside each bar is decided/total probes for that category.

Expired-figure failure
58%36/36
Hallucinated-fact failure
57%30/30
Product-recommendation failure
17%64/64
Outcome-promise failure
61%36/36
Missing-caveat failure
15%46/46
Referenceability failure
18%34/34
Completeness-gap failure
58%38/38
Bias-exploitation failure
28%32/32
Emotion-manipulation failure
0%32/32
Understanding-check failure
92%36/36
Information-overload failure
100%36/36
Missing-friction failure
12%34/34
Vulnerability-tailoring failure
0%32/32
Inappropriate-urgency failure
25%32/32
Naming a bias helpfully
3%32/32

← Back to the leaderboard