Benchmarks

Sentiqra vs the Competition

Side-by-side comparison of Top-3 Hit Rate, taxonomy depth, determinism, audit trails, and cost across the major content moderation APIs.

Benchmarks by Service

Trust & Safety

210 indicators · 21 harm groups · 16 languages

95.7–100% Production Top-3

Cybersecurity

200 indicators · 6 threat groups · 15 languages

85.3–96.6% Production Top-3

Discrimination Analysis

200 indicators · 6 discrimination groups · 16 languages

88–98.7% Production Top-3

Dating Safety

200 indicators · 5 groups · 40 languages

Benchmark pending

Fraud & Scam Detection

200 indicators · 4 groups · 39 languages

Benchmark pending

Crisis Intelligence

650 indicators · 7 analysis groups · 21 languages

86.7–96.7% Production Top-3

Disinformation Analysis

200 indicators · 6 manipulation groups · 16 languages

87.3–100% Production Top-3

Market Intelligence

200 indicators · 20 event groups · 8 languages

85.4–93.2% Production Top-3

Insurance Intelligence

200 indicators · 4 groups · 40 languages

Benchmark pending

Emotion Analysis

20 emotions · 20 emotion groups · 16 languages

87.3–95.2% Production Top-3

Customer Service Intelligence

200 indicators · 6 support groups · 18 languages

87.3–96.7% Production Top-3

Wellness Intelligence

200 indicators · 4 groups · 41 languages

Benchmark pending

Note: Sentiqra results are generated reproducibly from version-controlled evaluation artifacts and distinguish ranked Top-3 from production-gated Top-3. Competitor claims carry a per-vendor verification date, model/API version, and source above; figures are self-reported by the respective vendors and re-verified at least every 90 days - comparisons older than that are marked under review rather than presented as current. We encourage you to run your own evaluation on your content.

Sentiqra Trust & Safety Results

Language	Ranked Top-3	Production Top-3
Englishbench-ts-2026-09-en	98.0%	98.0%
Germanbench-ts-2026-09-de	98.0%	98.0%
Frenchbench-ts-2026-09-fr	98.0%	98.0%

Production Top-3 applies the production quality gate used by the API. Ranked Top-3 is shown separately for comparison. Representative languages shown - the complete 16-language matrix and sample counts are available to customers and evaluators after sign-in. See benchmark methodology for test details.

These are hit rates over a labeled benchmark set, not alert precision on your traffic: the share of flags that are true positives depends on how prevalent the target condition is in your content. Translate your own measured rates into expected alerts and review workload with sentiqra-prevalence.py. Negative-case evidence: the held-out decision-quality evaluation reports false-positive rates, abstention, and weak spots with uncertainty intervals.

Feature Comparison

Feature	Sentiqra T&S	Perspective	OpenAI Mod	Azure CS	GPT-4o (prompted)
Competitor claims		Verified Sep 26, 2026 Perspective API - 6 production attributes with per-attribute probability scores; sunsetting: service ends December 31, 2026; no new usage requests after February 2026 Perspective API docs (attributes + FAQ)	Verified Sep 26, 2026 omni-moderation-latest (13 categories with per-category category_scores, text + image) OpenAI Moderation guide	Verified Sep 21, 2026 Azure AI Content Safety text model - 4 categories, severity scale 0-7 Microsoft Learn: harm categories	Verified Sep 21, 2026 gpt-4o-2024-05-13 - API retirement scheduled October 23, 2026 OpenAI deprecations
Type	Deterministic, versioned analysis engine	ML classifier (attribute scores)	GPT-based classifier	Multi-modal classifier	General-purpose LLM + prompt
Languages	16	~18 (varies by attribute)	100+ (quality varies)	100+ (quality varies)	95+ (unverified)
Indicators	210	6 production attributes + experimental	13 categories	4 categories (hate, sexual, violence, self-harm) + severity levels	None (free-form labels)
Avg Top-3 Hit Rate (T&S generated eval, all 16 languages)	97.9%	Not published	85% (self-reported)	90% (self-reported)	Not comparable
Latency p50	See performance dashboard	~200ms	~300ms	~250ms	~1-3s (variable)
Latency p95	See performance dashboard	~500ms	~800ms	~600ms	Variable
Cost / 1K calls	Included on all plans; overage $0.90-2.50/1K	Free (rate-limited)	Free for OpenAI API customers	Per-1K-records billing (see Azure pricing)	~$2.50/1M input tokens + prompt maintenance
Deterministic output	Yes	Yes	No	Yes	No
Replayable receipt (audit)	Yes (reproducible)	No	No	No	No
Per-Indicator Signal Score	Yes (signal-match strength)	No (single score)	Binary per-category	Severity level	No
Custom Taxonomy	Yes	No	No	No	Prompt-defined only

Competitor Analysis

Google Perspective API

Per-attribute probability scores (TOXICITY, SEVERE_TOXICITY, IDENTITY_ATTACK, INSULT, PROFANITY, THREAT). No published per-language Top-3 Hit Rate benchmarks.

Strengths

Free
Per-attribute probability scores
Simple integration

Weaknesses

Small fixed attribute set (no 200-indicator taxonomy)
No published per-language benchmark evidence
No audit trail
No custom taxonomy
Black-box model

OpenAI Moderation API

Per-category flags and 0-1 category_scores returned. Self-reported accuracy on OpenAI's test set; no per-language breakdown published. omni-moderation-latest covers 13 categories and accepts text + image.

Strengths

Free for OpenAI customers
13 categories with per-category scores
Wide language coverage

Weaknesses

Non-deterministic (LLM-based)
No custom taxonomy
No audit trail
Broad categories (no 200-indicator granularity)
No per-language benchmark evidence

Azure AI Content Safety

Self-reported accuracy on Microsoft's test set; no per-language breakdown published. Categories are broad with severity levels (0-7 scale for text).

Strengths

Azure ecosystem integration
Severity levels
Multi-modal (text + image)

Weaknesses

4 categories only (no 200-indicator taxonomy)
No custom taxonomy
No audit trail
No per-language benchmark evidence

GPT-4o (prompted classifier)

No published classification benchmark - accuracy depends on the prompt and model version, and changes when the model updates.

Strengths

Flexible - any label set via prompt
Reasoning on edge cases

Weaknesses

Non-deterministic
Behavior changes on model updates
gpt-4o-2024-05-13 API snapshot retires October 23, 2026
No fixed taxonomy
No per-signal score
No reproducible audit trail

Key Differences

Precision Sentiqra publishes per-language ranked and production Top-3 results. Competitors do not publish equivalent per-language Top-3 evidence.

Taxonomy depth Sentiqra Trust & Safety: 210 indicators in 21 groups. Perspective: 6 production attributes. OpenAI: 13 categories. Azure: 4 categories.

Determinism Sentiqra: same input against the same pinned configuration always returns the same output. Perspective/Azure: deterministic. OpenAI/GPT-4o: non-deterministic.

Audit trail Sentiqra: full reproducible audit trail (config ID + fingerprint + signature). All competitors: no audit trail.

Per-indicator signal score Sentiqra: per-indicator signal-match scores. OpenAI: per-category scores (category_scores). Perspective: per-attribute probability scores. Azure: severity levels.

Custom taxonomy Sentiqra: custom taxonomy supported. All competitors: fixed taxonomy.

Cost Sentiqra: runs included on all plans; overage $0.90-2.50/1K by plan. Perspective: free. OpenAI: free for customers. Azure: billed per 1K text records. GPT-4o: per-token pricing plus prompt maintenance.

Benchmark Results

Per-language Top-3 Hit Rate

Benchmark Methodology

How we measure Top-3 Hit Rate

vs LLM Moderation

Sentiqra vs Perspective, GPT-4o

Performance Dashboard

Latency, uptime, throughput snapshot

Playground

Test detection live - no API key

Pricing

From Free to Enterprise

Quickstart Guide

First API call in under 5 minutes

Legal Disclaimer

Benchmark methodology claims