The LPDDR5 memory is listed with a 16GB maximum; no RAM upgrade is assumed.
Listing evidence: 2026-09-25. Amazon.com base price visible in an owner-supplied screenshot reviewed 25 Sep 2026; capture time not supplied. Dated reference, not a live quote. Excludes card signup offers, coupons, tax and shipping.
Fixed 24GB memory and internal storage limit model size; no CUDA support; Amazon main-offer price and stock were not verified in this run.
Listing evidence: 2026-10-03. Apple US launch price announced 22 September 2026: USD 1,699. Dated manufacturer reference, not an Amazon offer; tax and shipping may differ.
Memory is soldered and shared with the operating system.
Listing evidence: 2026-09-24. GMKtec, EVO-X2 / 64GB / 1TB, recorded 22 Sep 2026; exact ASIN capacity identified 24 Sep. This is a dated reference, not a current Amazon price; tax, shipping and bundle contents may differ.
Weight size is not tested compatibility or speed. Check runtime support and memory for context before buying. Prices are dated references. As an Amazon Associate I earn from qualifying purchases.
How these options are selected
Up to three platforms, with the lowest dated reference price on each. Unknown prices follow known prices. Each memory pool keeps a reserve: 30% (minimum 2 GB) for system or unified memory; 20% (minimum 1 GB) for dedicated VRAM. System RAM is for CPU inference. This check does not establish GPU allocation, software support or context capacity.
Model file and details
Provider
Liquid AI
Total parameters
1,918,181,888
Exact file size
1,099,627,808 bytes
Quantization / precision
Q4_K_M
Weight repository
LiquidAI/LFM3-VL-2B-GGUF
Original model repository
LiquidAI/LFM3-VL-2B
Weight revision
b36554cfd35854b06429900131730cd80f092cdb
Model revision
2bce59575de21cc5029fcd16848a05c88688be6f
Runtime notes
Confirm model architecture and quantization support in your runner.
Recorded release date
2026-10-27
Date evidence
Liquid AI LFM3 family announcement dated 27 October 2026; official source and GGUF model cards reviewed on 28 October. Exact source parameter count includes modality encoders; the language GGUF metadata count excludes separate encoders.
These scores were compiled from the linked sources. These published scores were not independently tested here. Different prompts, precision, thinking budgets and tools can change results.
DocVQA val · ANLS
96.7 / 100
Publisher-reported DocVQA validation ANLS normalized to 0–100. Document reading, not general chat quality.
Benchmark efficiency: 87.939 published score points per GB of selected weights. This ratio is not an inference-speed measure; check the test settings to see whether the score covers the selected quantization.
Test settings: Liquid AI source-checkpoint evaluation with vLLM 0.27.1, reasoning enabled and recommended generation parameters. DocVQA validation ANLS is multiplied by 100; ChartQA uses the test split. No separate quality score for Q4_K_M is given.
Publisher source-checkpoint result, not an independent evaluation or a measurement of the selected quantized file. No repeated-run confidence interval is supplied. File size covers language weights only; it is not runtime RAM. Allow additional memory for context, runtime and media. LFM Open License v1.0 applies; commercial use at or above the defined USD 10 million annual-revenue threshold needs a separate license. The separate Q8_0 projector is 203,031,872 bytes; language plus projector is 1,302,659,680 bytes before runtime/context overhead. Projector: https://huggingface.co/LiquidAI/LFM3-VL-2B-GGUF/blob/b36554cfd35854b06429900131730cd80f092cdb/mmproj-LFM3-VL-2B-Q8_0.gguf. Text, images and video input; 131,072-token source limit. Media support differs by app: the official card lists image/video in Llama.app, image input in LM Studio and Ollama. Use a LFM3-capable engine; source context does not establish practical device capacity.
Publisher-reported ChartQA test accuracy. Read each model protocol for reasoning and generation settings.
Benchmark efficiency: 84.301 published score points per GB of selected weights. This ratio is not an inference-speed measure; check the test settings to see whether the score covers the selected quantization.
Test settings: Liquid AI source-checkpoint evaluation with vLLM 0.27.1, reasoning enabled and recommended generation parameters. DocVQA validation ANLS is multiplied by 100; ChartQA uses the test split. No separate quality score for Q4_K_M is given.
Publisher source-checkpoint result, not an independent evaluation or a measurement of the selected quantized file. No repeated-run confidence interval is supplied. File size covers language weights only; it is not runtime RAM. Allow additional memory for context, runtime and media. LFM Open License v1.0 applies; commercial use at or above the defined USD 10 million annual-revenue threshold needs a separate license. The separate Q8_0 projector is 203,031,872 bytes; language plus projector is 1,302,659,680 bytes before runtime/context overhead. Projector: https://huggingface.co/LiquidAI/LFM3-VL-2B-GGUF/blob/b36554cfd35854b06429900131730cd80f092cdb/mmproj-LFM3-VL-2B-Q8_0.gguf. Text, images and video input; 131,072-token source limit. Media support differs by app: the official card lists image/video in Llama.app, image input in LM Studio and Ollama. Use a LFM3-capable engine; source context does not establish practical device capacity.
Context and version evidence: The repository and file name identify the selected model. Revision hashes are shown above when recorded. Tested context length, full hardware configuration and run date are not consistently recorded in this snapshot. Only settings explicitly stated in each benchmark record are known.
The 1.1 GB figure covers model weights in this file. It excludes the context cache, runtime, operating system and any separate multimodal files. It is not a measured RAM requirement or a token-generation speed.
Confirm model architecture and quantization support in your runner. The selected weights use 1.1 GB; this excludes context cache, runtime buffers and other programs.
These are the first listed tiers that pass our reserve policy, not measured minimum requirements. System RAM and VRAM remain separate. Use the memory tool to change the reserve for your context and runner.