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
Claim Continuity Classification
Classifies whether a previously verified claim remains materially continuous, changes, becomes unsupported, or is superseded under a new subject snapshot and evidence bundle. Illustrative uses include …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| GPT-6 Luna ★ | 92% | RANKED | best value |
| Xiaomi MiMo V2.6 Flash | 95% | MEDIUM | 4.8x |
| GPT-6 Sol | 96% | RANKED | 16x |
| Claude Sonnet 5.5 best | 100% | HIGH | 30x |
| Meta Muse Spark 1.3 | 94% | RANKED | 38x |
| Moonshot Kimi K3 | 99% | HIGH | 194x |
| Claude Opus 5.5 | 99% | HIGH | 289x |
Public Engagement Reply Review
Reviews a caller-supplied public response against its source message, allowed evidence, forbidden claims, operator instructions, channel rules, and review policy, returning an advisory readiness …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| GLM-5.3 Flash ★ best | 100% | MEDIUM | best value |
| Tencent Hy4 Preview | 94% | MEDIUM | 4.8x |
Capability Category Matching
Assigns a task or capability definition to a supplied category taxonomy or proposes a new category when the taxonomy has no adequate semantic match. Illustrative uses include categorizing software …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| Xiaomi MiMo V2.6 Flash ★ | 99% | MEDIUM | best value |
| Claude Haiku 5.5 | 96% | MEDIUM | 2.4x |
| Xiaomi MiMo V2.6 Pro | 100% | MEDIUM | 2.9x |
| DeepSeek V4.1 Flash best | 100% | HIGH | 5.2x |
| Claude Sonnet 5.5 | 100% | MEDIUM | 35x |
| Claude Opus 5.5 | 99% | MEDIUM | 75x |
Generated Content Relevance Scoring
Scores every item in a supplied set of generated or intermediate content against a target context, objective, or specification, returning a calibrated item-level relevance assessment. One current …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| DeepSeek V4.1 Flash ★ | 96% | MEDIUM | best value |
| Xiaomi MiMo V2.6 Pro | 94% | MEDIUM | 1.4x |
| GPT-6 Luna | 94% | MEDIUM | 2.4x |
| Gemini 3.5 Flash Lite | 91% | MEDIUM | 3.8x |
| Qwen 3.7 Plus best | 100% | HIGH | 6.1x |
| GPT-6 Sol | 93% | MEDIUM | 8.9x |
| Thinking Machines Inkling Small | 92% | MEDIUM | 9x |
| Grok 4.7 | 98% | HIGH | 21x |
| Claude Opus 5.5 | 95% | MEDIUM | 37x |
Authoritative Source Selection
Recommends authoritative, accessible, and maintainable sources for researching a subject, focus, regions, and optional analysis context. Illustrative uses include selecting sources for vendor …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| GLM-5.3 Flash ★ | 94% | HIGH | best value |
| Meta Muse Spark 1.3 | 95% | RANKED | 2x |
| Tencent Hy4 Preview | 94% | MEDIUM | 2.2x |
| Claude Opus 5.5 | 97% | RANKED | 3.6x |
| Grok 4.7 | 96% | HIGH | 3.8x |
| Moonshot Kimi K3 | 93% | MEDIUM | 4.9x |
| GLM-5.3 best | 100% | RANKED | 10x |
Community Policy Evaluation
Determines whether supplied community rules permit a proposed action under caller-defined channel and decision policies, returning a conservative decision and evidence. Illustrative uses include …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| MiniMax M3 ★ best | 100% | HIGH | best value |
Language Identification
Identifies the primary language of supplied text and returns its normalized language code in the configured response schema. Illustrative uses include routing multilingual support tickets, contracts, …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| Xiaomi MiMo V2.6 Flash ★ best | 100% | MEDIUM | best value |
| GPT-6 Luna | 99% | RANKED | 1.4x |
| Gemini 3.5 Flash Lite | 98% | RANKED | 2.6x |
| NVIDIA Nemotron-3 Nano 30B-A3B | 98% | RANKED | 2.8x |
| NVIDIA Nemotron-3 Super 120B | 98% | RANKED | 3.5x |
| Xiaomi MiMo V2.6 Pro | 99% | RANKED | 6.5x |
| MiniMax M3 | 98% | RANKED | 6.7x |
| NVIDIA Nemotron 3.5 Lightning | 95% | MEDIUM | 6.9x |
| Thinking Machines Inkling Small | 97% | RANKED | 8.4x |
| Tencent Hy3 | 98% | RANKED | 11x |
| Qwen 3.8 Flash | 97% | RANKED | 12x |
| Gemini 3.8 Flash | 98% | RANKED | 13x |
| GLM-5.3 Flash | 96% | HIGH | 16x |
| GPT-6 Sol | 95% | MEDIUM | 22x |
| NVIDIA Nemotron-3 Ultra 550B | 98% | RANKED | 26x |
| Qwen 3.7 Plus | 98% | RANKED | 38x |
| Claude Opus 5.5 | 97% | MEDIUM | 63x |
| Thinking Machines Inkling | 96% | MEDIUM | 68x |
| Tencent Hy4 Preview | 98% | HIGH | 76x |
| Qwen 3.8 Max | 99% | RANKED | 83x |
| Moonshot Kimi K3 | 99% | RANKED | 95x |
| GLM-5.3 | 98% | HIGH | 123x |
| Meta Muse Spark 1.3 | 97% | HIGH | 129x |
Persona Policy Eligibility Check
Determines whether a proposed persona or identity use satisfies a caller-supplied eligibility policy and evidence standard. Illustrative uses include checking a branded virtual spokesperson, …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| Claude Haiku 5.5 ★ | 94% | HIGH | best value |
| Xiaomi MiMo V2.6 Pro | 95% | MEDIUM | 1.4x |
| DeepSeek V4.1 Flash | 97% | HIGH | 2.4x |
| Gemini 3.8 Flash | 94% | HIGH | 8.6x |
| GLM-5.3 Flash | 99% | MEDIUM | 15x |
| Claude Opus 5.5 | 100% | RANKED | 35x |
| Meta Muse Spark 1.3 | 97% | HIGH | 36x |
| Thinking Machines Inkling | 92% | MEDIUM | 43x |
| Moonshot Kimi K3 best | 100% | HIGH | 58x |
| Grok 4.7 | 92% | MEDIUM | 72x |
| Tencent Hy4 Preview | 97% | MEDIUM | 82x |
| GLM-5.3 | 100% | MEDIUM | 175x |
Profile Pool Matching
Matches a source object to the most suitable candidates in a supplied profile pool using caller-defined dimensions, eligibility rules, thresholds, and creation policy. When permitted, it may propose a …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| Xiaomi MiMo V2.6 Flash ★ | 94% | MEDIUM | best value |
| Claude Haiku 5.5 | 95% | MEDIUM | 1.9x |
| Qwen 3.7 Plus | 90% | MEDIUM | 9.7x |
| Thinking Machines Inkling Small | 90% | MEDIUM | 13x |
| GLM-5.3 Flash | 99% | MEDIUM | 14x |
| Claude Sonnet 5.5 | 95% | MEDIUM | 35x |
| Meta Muse Spark 1.3 | 98% | MEDIUM | 46x |
| Tencent Hy4 Preview best | 100% | MEDIUM | 48x |
| Thinking Machines Inkling | 93% | MEDIUM | 51x |
| Claude Opus 5.5 | 95% | MEDIUM | 71x |
| Grok 4.7 | 90% | MEDIUM | 100x |
| GLM-5.3 | 96% | MEDIUM | 157x |
Engagement Opportunity Triage
Classifies whether an incoming public message merits a response, identifies risk, and proposes a defensible angle using supplied themes, evidence boundaries, channel context, and policy. Illustrative …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| Xiaomi MiMo V2.6 Flash ★ | 94% | MEDIUM | best value |
| Claude Haiku 5.5 | 96% | MEDIUM | 2.7x |
| Gemini 3.5 Flash Lite | 93% | HIGH | 4.6x |
| Xiaomi MiMo V2.6 Pro | 98% | MEDIUM | 5.3x |
| MiniMax M3 | 97% | RANKED | 5.9x |
| NVIDIA Nemotron-3 Super 120B | 96% | RANKED | 9.3x |
| Tencent Hy3 | 96% | RANKED | 11x |
| Qwen 3.8 Flash | 92% | MEDIUM | 16x |
| Gemini 3.8 Flash | 96% | RANKED | 16x |
| GPT-6 Sol | 91% | HIGH | 21x |
| Thinking Machines Inkling Small | 99% | RANKED | 24x |
| GLM-5.3 Flash | 95% | MEDIUM | 25x |
| Qwen 3.7 Plus | 98% | RANKED | 25x |
| NVIDIA Nemotron-3 Ultra 550B | 98% | RANKED | 35x |
| Claude Sonnet 5.5 | 96% | MEDIUM | 37x |
| Claude Opus 5.5 | 96% | MEDIUM | 94x |
| Thinking Machines Inkling best | 100% | RANKED | 124x |
| Qwen 3.8 Max | 95% | RANKED | 130x |
| Grok 4.7 | 96% | MEDIUM | 137x |
| Moonshot Kimi K3 | 100% | HIGH | 166x |
| GLM-5.3 | 98% | RANKED | 243x |
Content Set Relevance Scoring
Scores every item in one supplied set of retrieved content for relevance to a query or analysis objective, returning a calibrated item-level score with concise evidence for each. Illustrative uses …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| DeepSeek V4.1 Flash ★ | 100% | MEDIUM | best value |
| Xiaomi MiMo V2.6 Pro best | 100% | MEDIUM | 1.4x |
| GPT-6 Luna | 90% | MEDIUM | 1.5x |
| Thinking Machines Inkling Small | 95% | MEDIUM | 5.2x |
| GPT-6 Sol | 97% | MEDIUM | 28x |
Social Post Portfolio Selection
Selects the best fixed-size portfolio of candidate social posts using quality, expected audience value, topical diversity, timeliness, and caller-supplied constraints. Illustrative uses include …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| GPT-6 Luna ★ | 98% | HIGH | best value |
| Tencent Hy3 | 91% | HIGH | 1.3x |
| Xiaomi MiMo V2.6 Flash | 100% | RANKED | 1.4x |
| Claude Haiku 5.5 | 99% | RANKED | 1.4x |
| Gemini 3.5 Flash Lite | 98% | HIGH | 1.9x |
| NVIDIA Nemotron-3 Nano 30B-A3B | 95% | MEDIUM | 2.1x |
| Xiaomi MiMo V2.6 Pro | 97% | HIGH | 3.5x |
| MiniMax M3 | 96% | RANKED | 4.1x |
| NVIDIA Nemotron-3 Super 120B | 93% | MEDIUM | 9.4x |
| NVIDIA Nemotron 3.5 Lightning | 92% | HIGH | 9.5x |
| NVIDIA Nemotron-3 Ultra 550B | 95% | MEDIUM | 13x |
| GPT-6 Sol | 100% | RANKED | 17x |
| Claude Sonnet 5.5 | 98% | HIGH | 19x |
| DeepSeek V4.1 Flash | 97% | RANKED | 19x |
| Qwen 3.7 Plus | 96% | RANKED | 23x |
| Gemini 3.8 Flash | 100% | RANKED | 23x |
| Thinking Machines Inkling Small | 100% | HIGH | 25x |
| GLM-5.3 Flash | 99% | RANKED | 27x |
| Claude Opus 5.5 | 98% | HIGH | 42x |
| Qwen 3.8 Flash | 94% | HIGH | 51x |
| Meta Muse Spark 1.3 best | 100% | RANKED | 94x |
| Grok 4.7 | 97% | HIGH | 131x |
| Thinking Machines Inkling | 98% | HIGH | 171x |
| Qwen 3.8 Max | 93% | RANKED | 180x |
| Moonshot Kimi K3 | 99% | HIGH | 184x |
| Tencent Hy4 Preview | 99% | RANKED | 206x |
| GLM-5.3 | 98% | RANKED | 213x |
Research Community Selection
Selects relevant, accessible communities as research sources for a subject, purpose, regions, and platform profile. Illustrative uses include finding developer forums, customer user groups, …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| DeepSeek V4.1 Flash ★ | 96% | HIGH | best value |
| Qwen 3.7 Plus | 98% | MEDIUM | 1.6x |
| Gemini 3.8 Flash | 90% | MEDIUM | 2.1x |
| GPT-6 Sol | 98% | HIGH | 2.6x |
| GLM-5.3 Flash | 97% | HIGH | 3.3x |
| Claude Sonnet 5.5 | 91% | MEDIUM | 5.9x |
| Meta Muse Spark 1.3 best | 100% | HIGH | 6.5x |
| Claude Opus 5.5 | 97% | HIGH | 6.6x |
| Grok 4.7 | 98% | HIGH | 15x |
| Qwen 3.8 Max | 92% | MEDIUM | 24x |
Content Domain Suggestion
Suggests one to three concise subject-domain labels for a content or analysis specification using its title, description, subject context, structure, and any existing domain vocabulary. Illustrative …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| Thinking Machines Inkling Small ★ best | 100% | MEDIUM | best value |
| Claude Opus 5.5 | 95% | MEDIUM | 4.2x |
Short Post Batch Relevance Scoring
Scores a batch of short-form posts against a supplied topic, query, or scoring context, returning an item-level relevance score for each post. It is intended for compact posts such as those from X, …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| MiniMax M3 ★ best | 100% | RANKED | best value |
| Xiaomi MiMo V2.6 Pro | 93% | MEDIUM | 3.5x |
| Gemini 3.8 Flash | 93% | MEDIUM | 8.6x |
| Claude Haiku 5.5 | 93% | MEDIUM | 8.6x |
| GLM-5.3 Flash | 93% | MEDIUM | 9.8x |
| GPT-6 Sol | 95% | MEDIUM | 18x |
| Meta Muse Spark 1.3 | 95% | MEDIUM | 29x |
| Moonshot Kimi K3 | 94% | MEDIUM | 63x |
| Claude Opus 5.5 | 91% | MEDIUM | 67x |
Topic Taxonomy Matching
Maps transient input topics to persistent taxonomy topics, reusing existing topics where semantically appropriate and proposing new ones only for durable coverage gaps. Illustrative uses include …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| Qwen 3.7 Plus ★ | 90% | MEDIUM | best value |
| Gemini 3.8 Flash best | 100% | MEDIUM | 1.8x |
| Meta Muse Spark 1.3 | 93% | MEDIUM | 4.9x |
| Moonshot Kimi K3 | 97% | MEDIUM | 14x |
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.