Model reference — Documents
Sentiment analysis judges how a piece of text feels — positive, negative, neutral, or a finer scale — and intent analysis judges what the writer wants: to buy, to cancel, to complain, to escalate to a regulator.
The two are usually run together on the same text, because a decision needs both. A calmly worded cancellation matters more than an angry question. The useful output is rarely a single label: it is a sentiment score, an intent, the confidence in each, and the sentence that drove the judgement, so a person can check it. Models range from small classifiers fitted to your own labels to language models that judge from a written definition of each intent.
Models in this group take text or whole documents as input: plain text, PDFs, scanned pages and office files. Each specification table gives three hardware tiers — Minimum, the smallest setup on which the model runs correctly; Medium, the usual production configuration; and High, a configuration sized for peak volume. The rate is the number of sample inputs processed per hour on that hardware. Use these rates for initial sizing. Before production, benchmark your own data to validate accuracy, latency, throughput and cost. The sample input here is one text of about 200 words scored for sentiment and intent.
Vendor: Meta
What it does: scores text as positive, negative or neutral, already fitted on large volumes of reviews and social messages. Accurate out of the box on short consumer text and cheap to run at scale.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3060 | RTX 4090 | 1× A100 80 GB |
| VRAM | 6 GB | 24 GB | 80 GB |
| vCPUs | 4 | 8 | 24 |
| RAM | 16 GB | 32 GB | 64 GB |
| Server | 1× RTX 3060 12 GB | 1× RTX 4090 24 GB | 1× A100 SXM 80 GB |
| Rate (texts/hour) | ≈ 40,000 | ≈ 160,000 | ≈ 450,000 |
Vendor: Microsoft
What it does: fitted to your own intent list from labelled examples. The most accurate option once you have a few hundred examples of each intent, and it handles longer text than review-tuned models.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3060 | RTX 4090 | 1× A100 80 GB |
| VRAM | 6 GB | 24 GB | 80 GB |
| vCPUs | 4 | 8 | 24 |
| RAM | 16 GB | 32 GB | 64 GB |
| Server | 1× RTX 3060 12 GB | 1× RTX 4090 24 GB | 1× A100 SXM 80 GB |
| Rate (texts/hour) | ≈ 30,000 | ≈ 120,000 | ≈ 340,000 |
Vendor: Meta
What it does: judges sentiment and intent across a hundred languages with one model, so a multinational support desk does not need a separate model per market.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3060 | RTX 4090 | 1× A100 80 GB |
| VRAM | 8 GB | 24 GB | 80 GB |
| vCPUs | 6 | 8 | 24 |
| RAM | 16 GB | 32 GB | 64 GB |
| Server | 1× RTX 3060 12 GB | 1× RTX 4090 24 GB | 1× A100 SXM 80 GB |
| Rate (texts/hour) | ≈ 26,000 | ≈ 100,000 | ≈ 300,000 |
Vendor: Hugging Face
What it does: fits a working intent classifier from a handful of examples per intent, which is how a new intent — a new product complaint, a new regulation — gets covered within a day.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3060 | RTX 4090 | 1× A100 80 GB |
| VRAM | 6 GB | 24 GB | 80 GB |
| vCPUs | 4 | 8 | 24 |
| RAM | 16 GB | 32 GB | 64 GB |
| Server | 1× RTX 3060 12 GB | 1× RTX 4090 24 GB | 1× A100 SXM 80 GB |
| Rate (texts/hour) | ≈ 28,000 | ≈ 110,000 | ≈ 320,000 |
Vendor: Meta
What it does: judges sentiment and intent from written definitions and returns the sentence that decided it, which makes its reasoning auditable. Also handles sarcasm and mixed messages better than small classifiers.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 3090 | RTX 4090 | 1× H100 80 GB |
| VRAM | 16 GB | 24 GB | 80 GB |
| vCPUs | 8 | 12 | 24 |
| RAM | 32 GB | 48 GB | 128 GB |
| Server | 1× RTX 3090 24 GB | 1× RTX 4090 24 GB | 1× H100 SXM 80 GB |
| Rate (texts/hour) | ≈ 2,000 | ≈ 6,000 | ≈ 20,000 |
Vendor: Mistral AI
What it does: the same instruction-driven judgement on long, formally written text such as written complaints and regulator correspondence, where tone is carried across paragraphs rather than in one phrase.
| Requirement | Minimum | Medium | High |
|---|---|---|---|
| GPU type | RTX 4090 | 1× L40S 48 GB | 1× H100 80 GB |
| VRAM | 24 GB | 48 GB | 80 GB |
| vCPUs | 12 | 16 | 32 |
| RAM | 48 GB | 64 GB | 128 GB |
| Server | 1× RTX 4090 24 GB | 1× L40S 48 GB | 1× H100 SXM 80 GB |
| Rate (texts/hour) | ≈ 1,300 | ≈ 3,800 | ≈ 11,000 |
Which model fits depends on the length and register of your text — a two-line review and a three-page complaint are different problems — and on whether the output triggers automatic action. Our consultants review your text and your intent list, then recommend a model and the confidence threshold above which its judgement can be acted on unattended.
At the start of a project we may run a short proof of concept on a sample of your own data, measuring the accuracy and the throughput the model actually achieves on your material. That replaces the estimates on this page with real figures, so the cost and the schedule for the full engagement are known before it is committed.
Share a representative sample, expected volume, latency target and deployment location for Sentiment / intent analysis. We benchmark the shortlisted models, recommend the lowest-cost GPU configuration that meets the target, and scale it from pilot capacity to a dedicated production cluster.