Best LLMs for Relevance, Classification & Matching
Semantic similarity judgment: does this thing belong in that bucket / match that target?
Semantic similarity judgment: does this thing belong in that bucket / match that target?
Task-by-task breakdown
Vetted News Site Selection
Identifies authoritative, accessible web sources for researching a subject within specified regions. Prioritises official, recognised, regularly maintained sources; filters out strict-paywall and …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| MiniMax M3 ★ | 90% | RANKED | best value |
| DeepSeek V4 Pro | 92% | MEDIUM | 1.3x |
| Qwen 3.6 Plus | 96% | HIGH | 2.7x |
| Meta Muse Spark 1.1 | 91% | HIGH | 5.7x |
| GPT-5.6 Sol best | 100% | HIGH | 11x |
| GPT-5.5 | 99% | MEDIUM | 21x |
Content Domain Suggestion
Picks 1-3 subject/vertical domain tags for an analysis template (e.g. 'Menopause', 'Femtech', 'Biotech') from its name, description, and chapter content. Names the WHAT, not the HOW — explicitly …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| Meta Muse Spark 1.1 ★ best | 100% | MEDIUM | best value |
X Post Relevance Scoring
Scores batched X-com posts against synthesis capability (x_post_relevance stage). Split from pooled relevance_scoring on 2026-05-17. GENERIC_RELEVANCE_SCORE_{SYSTEM,USER}_PROMPT, batched input (≥20k …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| MiniMax M3 ★ | 98% | RANKED | best value |
| Gemini 3.5 Flash best | 100% | MEDIUM | 2.6x |
| Meta Muse Spark 1.1 | 94% | MEDIUM | 5.6x |
| Grok 4.5 | 92% | MEDIUM | 7x |
Engagement Reply Review
Independent quality review of a drafted engagement reply.
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| Gemini 3.1 Flash Lite ★ | 99% | RANKED | best value |
| Qwen 3.5 Flash | 98% | RANKED | 8.5x |
| Claude Sonnet 5 | 91% | MEDIUM | 30x |
| Qwen 3.6 Plus best | 100% | RANKED | 38x |
| Claude Opus 4.8 | 99% | RANKED | 58x |
| Kimi K2.6 | 96% | MEDIUM | 83x |
Subreddit Quality Vetting
Decides whether a subreddit permits automated posting via the Reddit API for substantive analytical content. Fails closed only on explicit prohibitions; topic restrictions, flair requirements, and …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| MiniMax M3 ★ | 90% | HIGH | best value |
| GPT-5.6 Luna best | 100% | MEDIUM | 2.5x |
| GPT-5.6 Terra | 98% | MEDIUM | 4.6x |
| Gemini 3.5 Flash | 92% | MEDIUM | 7.3x |
| Claude Sonnet 5 | 98% | HIGH | 7.7x |
| GPT-5.6 Sol | 98% | MEDIUM | 9.8x |
Social Post Relevance Scoring
Scores RetrievedContent against synthesis capability description (stage 40 relevance_analysis). Split from pooled relevance_scoring on 2026-05-17 to remove inter-family σ inflation. …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| Gemini 3.5 Flash ★ best | 100% | MEDIUM | best value |
| Grok 4.5 | 91% | HIGH | 1.3x |
Author Living-Person Safety Check
Postmortem-publicity-rights safety check for AI author personas. Determines whether the real figure behind a persona name has been deceased for ≥100 years (the threshold safely clears CA §3344.1, TN …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| DeepSeek V4 Flash ★ | 96% | MEDIUM | best value |
| DeepSeek V4 Pro best | 100% | RANKED | 5.8x |
| Qwen 3.6 Plus | 93% | MEDIUM | 11x |
| GPT-5.6 Terra | 95% | MEDIUM | 12x |
| Gemini 3.1 Pro Preview | 94% | HIGH | 16x |
| Gemini 3.5 Flash | 96% | RANKED | 18x |
| Kimi K2.6 | 96% | MEDIUM | 23x |
| GPT-5.6 Sol | 97% | HIGH | 23x |
| Meta Muse Spark 1.1 | 98% | HIGH | 25x |
| Grok 4.5 | 93% | RANKED | 43x |
| GPT-5.5 | 98% | MEDIUM | 59x |
Language Detection
Identifies the primary language of a text snippet. Returns only the two-letter ISO 639-1 code (e.g. 'en', 'es', 'zh'). Used upstream of the translation pipeline.
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| NVIDIA Nemotron-3 Nano 30B-A3B ★ | 95% | MEDIUM | best value |
| Gemini 3.1 Flash Lite | 99% | RANKED | 1.4x |
| DeepSeek V4 Flash | 98% | RANKED | 1.4x |
| GPT-5.4 Nano | 99% | RANKED | 1.5x |
| GPT-5.4 Mini | 99% | RANKED | 2.7x |
| NVIDIA Nemotron-3 Super 120B | 98% | HIGH | 3.1x |
| Tencent Hy3 | 98% | RANKED | 3.8x |
| MiniMax M3 | 99% | RANKED | 5.8x |
| DeepSeek V4 Pro | 98% | RANKED | 6.1x |
| GPT-5.6 Luna | 99% | RANKED | 7.2x |
| NVIDIA Nemotron-3 Ultra 550B | 99% | RANKED | 14x |
| Claude Haiku 4.5 | 99% | RANKED | 16x |
| Qwen 3.5 Flash | 99% | RANKED | 17x |
| GPT-5.6 Terra | 99% | RANKED | 17x |
| Gemini 3.5 Flash | 99% | RANKED | 35x |
| Claude Sonnet 5 best | 100% | RANKED | 35x |
| GPT-5.5 | 99% | RANKED | 37x |
| GPT-5.6 Sol | 99% | RANKED | 38x |
| Qwen 3.7 Plus | 99% | RANKED | 39x |
| Gemini 3.1 Pro Preview | 99% | RANKED | 48x |
| Qwen 3.6 Plus | 99% | RANKED | 50x |
| Claude Sonnet 4.6 | 99% | RANKED | 52x |
| Qwen 3.6 Flash | 99% | RANKED | 66x |
| Meta Muse Spark 1.1 | 98% | HIGH | 76x |
| Claude Opus 4.8 | 99% | RANKED | 86x |
| Grok 4.5 | 99% | RANKED | 89x |
| Kimi K2.6 | 98% | RANKED | 143x |
Topic Report Relevance Scoring
Scores TOPIC_REPORT PartialSyntheses against analysis template (stage 132) and report chapters (stage 134). Split from pooled relevance_scoring on 2026-05-17. …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| Qwen 3.6 Flash ★ | 91% | MEDIUM | best value |
| Gemini 3.5 Flash best | 100% | MEDIUM | 1.3x |
| Qwen 3.7 Plus | 97% | MEDIUM | 1.9x |
| GPT-5.6 Terra | 97% | MEDIUM | 3.6x |
| GPT-5.5 | 91% | HIGH | 4.3x |
| Grok 4.5 | 93% | RANKED | 4.9x |
| Meta Muse Spark 1.1 | 96% | MEDIUM | 5.1x |
| GPT-5.6 Sol | 100% | MEDIUM | 7.4x |
Subreddit Selection for Research
Picks relevant, active subreddits for researching a subject. Balances large communities (more content) with niche ones (more focused). Filters for accessibility (public, not quarantined) and quality …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| GPT-5.6 Terra ★ | 90% | MEDIUM | best value |
| GPT-5.6 Sol best | 100% | MEDIUM | 3.3x |
Topic Grouping and Client Matching
Groups workflow topics (per-article) under broader client topics (persistent categories) — either existing or new. Multiple workflow topics can and should share one client topic when they cover …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| DeepSeek V4 Flash ★ | 95% | RANKED | best value |
| DeepSeek V4 Pro | 94% | RANKED | 2.4x |
| GPT-5.6 Luna | 93% | MEDIUM | 2.7x |
| Qwen 3.7 Plus | 95% | HIGH | 4.3x |
| Qwen 3.6 Plus | 91% | HIGH | 4.6x |
| Gemini 3.1 Pro Preview | 93% | RANKED | 6.2x |
| Kimi K2.6 | 92% | HIGH | 10x |
| GPT-5.6 Terra | 98% | HIGH | 11x |
| Gemini 3.5 Flash | 99% | RANKED | 12x |
| Grok 4.5 | 98% | RANKED | 14x |
| Meta Muse Spark 1.1 | 96% | HIGH | 14x |
| GPT-5.6 Sol best | 100% | HIGH | 16x |
| GPT-5.5 | 94% | MEDIUM | 21x |
Engagement Triage
Decide engage/ignore + risk + angle for one social post.
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| DeepSeek V4 Flash ★ | 92% | HIGH | best value |
| GPT-5.4 Nano | 94% | RANKED | 1.6x |
| Gemini 3.1 Flash Lite | 94% | RANKED | 1.6x |
| Tencent Hy3 | 94% | MEDIUM | 1.8x |
| GPT-5.4 Mini | 93% | HIGH | 2.8x |
| NVIDIA Nemotron-3 Super 120B | 93% | MEDIUM | 3.8x |
| MiniMax M3 | 98% | RANKED | 4.7x |
| GPT-5.6 Luna | 97% | HIGH | 8.1x |
| Qwen 3.5 Flash | 94% | RANKED | 9.1x |
| Claude Haiku 4.5 | 98% | RANKED | 11x |
| GPT-5.6 Terra | 98% | RANKED | 15x |
| Qwen 3.7 Plus | 98% | RANKED | 17x |
| Gemini 3.5 Flash | 96% | RANKED | 23x |
| Gemini 3.1 Pro Preview | 95% | RANKED | 26x |
| Claude Sonnet 5 | 97% | RANKED | 28x |
| NVIDIA Nemotron-3 Ultra 550B | 96% | HIGH | 34x |
| Claude Sonnet 4.6 best | 100% | RANKED | 34x |
| GPT-5.6 Sol | 93% | MEDIUM | 38x |
| Qwen 3.6 Plus | 95% | HIGH | 43x |
| Grok 4.5 | 98% | RANKED | 46x |
| GPT-5.5 | 95% | RANKED | 55x |
| Meta Muse Spark 1.1 | 99% | HIGH | 57x |
| Claude Opus 4.8 | 98% | RANKED | 64x |
| Kimi K2.6 | 97% | RANKED | 77x |
X Post Selection
Picks the best N X.com posts to publish within a daily budget. Ranks by engagement potential, topic diversity (avoid bunching), content quality, and timeliness; returns only the chosen post IDs.
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| DeepSeek V4 Flash ★ | 100% | RANKED | best value |
| Gemini 3.1 Flash Lite | 97% | HIGH | 1.6x |
| GPT-5.4 Mini | 93% | MEDIUM | 2.1x |
| Qwen 3.5 Flash | 97% | HIGH | 2.7x |
| MiniMax M3 | 96% | RANKED | 4.7x |
| DeepSeek V4 Pro | 96% | HIGH | 4.9x |
| GPT-5.6 Luna | 96% | HIGH | 5.1x |
| Tencent Hy3 | 93% | HIGH | 6x |
| Qwen 3.6 Plus | 97% | RANKED | 7.5x |
| Claude Haiku 4.5 | 92% | MEDIUM | 7.6x |
| GPT-5.6 Terra | 98% | RANKED | 11x |
| Gemini 3.1 Pro Preview | 97% | HIGH | 12x |
| Kimi K2.6 | 99% | RANKED | 17x |
| Claude Sonnet 5 | 95% | RANKED | 20x |
| Claude Sonnet 4.6 | 100% | HIGH | 23x |
| Qwen 3.7 Plus | 95% | RANKED | 24x |
| GPT-5.5 best | 100% | RANKED | 25x |
| GPT-5.6 Sol | 97% | RANKED | 27x |
| Qwen 3.6 Flash | 98% | RANKED | 33x |
| Gemini 3.5 Flash | 97% | RANKED | 37x |
| Grok 4.5 | 95% | RANKED | 89x |
| Meta Muse Spark 1.1 | 96% | HIGH | 97x |
Author Matching
Matches content to fictional authors or creates new author personas
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| DeepSeek V4 Pro ★ | 93% | HIGH | best value |
| Qwen 3.6 Plus | 90% | MEDIUM | 1.4x |
| Gemini 3.1 Pro Preview | 93% | HIGH | 1.8x |
| Gemini 3.5 Flash | 92% | MEDIUM | 2.6x |
| Kimi K2.6 | 93% | HIGH | 3.5x |
| Claude Sonnet 4.6 | 91% | MEDIUM | 3.6x |
| Grok 4.5 best | 100% | MEDIUM | 5.2x |
| GPT-5.5 | 93% | HIGH | 6.4x |
Confidence — how sure we are about the quality score (more judgments + more agreement = higher confidence): RANKED many independent judges scored this model's outputs and their agreement is very high (most confident) — HIGH many judges have scored it and they mostly agree (well-pinned) — MEDIUM enough judges have weighed in to publish, but they disagree more than we'd like (treat with a small grain of salt). LOW-confidence cells are hidden everywhere on the site. See the methodology for the exact thresholds.