⟳
Evidence freshness sweep
ComplianceScan an evidence inventory and flag what is stale, expired, or due for re-collection before the audit window — with the cadence each item needs.
The routing recipe
Job typeStructured extractionPolicyCheapest that passes
routes to
GPT-OSS 20B groq
groq:openai/gpt-oss-20b
93.1
match score
80.9
task fit
100
cost
96.4
speed
Why this model from benchmarks & capabilities
- "Structured extraction" leans hardest on instruction; GPT-OSS 20B scores 78/100 there — #23 of 40.
- Strength on this class of work shows up on IFEval, IFBench (instruction following), where OpenAI (open-weight) models are competitive.
- Cost: free tier. Speed: ~1000 tok/s on Groq LPU.
- Chosen over Qwen3 Coder 480B-A35B Instruct for faster.
| Capability the job needs | Weight | GPT-OSS 20B | Catalog rank |
|---|---|---|---|
| instruction | 1.00 | #23 of 40 | |
| tool_use | 0.60 | #25 of 40 | |
| speed | 0.60 | #1 of 40 |
Relevant benchmarksIFEval IFBench (instruction following)
EconomicsCost: free tier.
Speed: ~1000 tok/s on Groq LPU. vs runner-upChosen over Qwen3 Coder 480B-A35B Instruct for faster. (Qwen3 Coder 480B-A35B Instruct, score 91.9)
Speed: ~1000 tok/s on Groq LPU. vs runner-upChosen over Qwen3 Coder 480B-A35B Instruct for faster. (Qwen3 Coder 480B-A35B Instruct, score 91.9)
Example result
Sample input ▾
Today 2026-06-26. Items: access review (2026-01-10, quarterly); pen test (2025-02-01, annual); background-check policy (2024-11-01, annual); vuln scan (2026-06-20, monthly).
EXPIRED — pen test (2025-02-01): annual, 4.5 mo overdue. Re-test now.
EXPIRED — background-check policy (2024-11-01): annual review 7 mo overdue.
STALE — access review (2026-01-10): quarterly, next due 2026-04-10 (overdue). Run Q2.
FRESH — vuln scan (2026-06-20): monthly, next due 2026-07-20.
Test it on your own data