Decision models · Ant Group (inclusionAI)

LemonDecider 4B

Task: Classifying and extracting
Returns: Probability or score
Precision tier: 80 / 100
Size: 4.5B
Max input: 8K tokens
License: apache-2.0
Released: 2026-10

What it decides

Returns: Probabilities for each option (up to 255 per question), yes/no probabilities or ordinal scores, for many named questions in one pass; it does not generate text.
Checks: Routing, intent, classification, NLI and fact checks, relevance, tool-call choice, hallucination, preference
Languages: English
Precision: 80 / 100 · Self-evaluated Decision Index (public 0.3) score of 60.03, with BANKING77 macro-F1 0.934, BFCL 0.953 and RAGTruth hallucination F1 0.843 (model card).
Clarity: Probability or score (15 of 15 clarity points)
How it runs: Classifier head · reads the input once and writes nothing, so it answers in milliseconds.
License: apache-2.0 (Open (Apache, MIT))

Source: model card on Hugging Face (opens in a new tab) · How decision models are rated

How to run it

Original weights · Hugging Face

Full-precision files from Ant Group (inclusionAI), for Transformers, Sentence Transformers, Text Embeddings Inference, Infinity or vLLM.

huggingface.co/inclusionAI/LemonDecider (opens in a new tab)

No sign-in needed.

Links checked 5 days ago against Hugging Face, Ollama and GitHub.

Apps that run LemonDecider 4B

Our pick for each kind of computer, assuming the model fits in its memory. Pick your own GPU or device in the Model Finder for an exact answer.

Mac (Apple silicon)

Sentence Transformers

A few lines of Python with Transformers or Sentence Transformers; easiest for scripts and notebooks.

Windows PC, NVIDIA GPU

Sentence Transformers

A few lines of Python with Transformers or Sentence Transformers; easiest for scripts and notebooks.

Windows PC, AMD Radeon

Sentence Transformers

A few lines of Python with Transformers or Sentence Transformers; easiest for scripts and notebooks.

Linux, NVIDIA GPU

Sentence Transformers

A few lines of Python with Transformers or Sentence Transformers; easiest for scripts and notebooks.

Linux server, 80 GB+ GPUs

Sentence Transformers

A few lines of Python with Transformers or Sentence Transformers; easiest for scripts and notebooks.

App	Type	Opens it as	Systems	Engine	Start
Sentence Transformers	Library	Original weights	macOSWindowsLinux	PyTorch	`python -c "from transformers import pipeline; print(pipeline('text-classification', model='inclusionAI/LemonDecider')('text'))"`

Which apps can open a model depends on the files published for it: GGUF for llama.cpp-based apps, MLX for Apple’s engine, the original weights for vLLM and SGLang. Compare every app.

Memory and speed

Each precision on an NVIDIA GeForce RTX 4090, for one decision on a 256-token input. The best hardware table below covers other systems and devices.

Format	Memory needed	Time per decision	Decisions per second	Quality	Download
`Q8_0`	6.3 GB	25 ms	40/s	80Great quality (A) · 80 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking.	All files (opens in a new tab)
`F16`	10.6 GB	28 ms	35/s	80Great quality (A) · 80 of 100: how capable the model is at F16, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking.	Original weights (opens in a new tab)

One request at a time. Servers that batch (Text Embeddings Inference, vLLM, Infinity) handle many more decisions per second.

Best hardware for LemonDecider 4B

The least expensive GPU in our catalog that runs it without offloading is the Intel Arc A580 (launch price $179) at Q8_0.

Quality	Hardware	Quant	Memory	Speed	Runs
80Great quality (A) · 80 of 100: how capable the model is at F16, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking.	AMD Radeon RX 7900 XTX	`F16`	10.8 GB / 24.0 GB	23 ms	Runs great
80Great quality (A) · 80 of 100: how capable the model is at F16, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking.	AMD Radeon RX 9070 XT	`F16`	10.8 GB / 16.0 GB	28 ms	Runs great
80Great quality (A) · 80 of 100: how capable the model is at F16, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking.	AMD Ryzen AI Max+ 395 (128 GB)	`F16`	10.8 GB / 96.0 GB	43 ms	Runs great
80Great quality (A) · 80 of 100: how capable the model is at F16, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking.	Apple M2 Ultra (192 GB)	`F16`	10.8 GB / 144 GB	90 ms	Runs great
80Great quality (A) · 80 of 100: how capable the model is at F16, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking.	Apple M3 Ultra (256 GB)	`F16`	10.8 GB / 192 GB	87 ms	Runs great
80Great quality (A) · 80 of 100: how capable the model is at F16, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking.	Apple M3 Ultra (512 GB)	`F16`	10.8 GB / 384 GB	87 ms	Runs great
80Great quality (A) · 80 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking.	Intel Arc B580	`Q8_0`	6.5 GB / 12.0 GB	88 ms	Runs great
80Great quality (A) · 80 of 100: how capable the model is at F16, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA GB10 (128 GB)	`F16`	10.8 GB / 115 GB	69 ms	Runs great
80Great quality (A) · 80 of 100: how capable the model is at F16, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA GeForce RTX 3090	`F16`	10.8 GB / 24.0 GB	61 ms	Runs great
80Great quality (A) · 80 of 100: how capable the model is at F16, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA GeForce RTX 4060 Ti 16GB	`F16`	10.8 GB / 16.0 GB	95 ms	Runs great

Check it on your own hardware

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.

Was this page helpful?