LensVLM 9B — Hardware Requirements & GPU Compatibility
Vision
LensVLM-9B is Apple's research vision-language model for efficient long-document understanding, built on a 9.4-billion-parameter Qwen3.5-9B backbone. Rather than feeding a document's full text into the context window, it scans compressed images of the text and uses learned tools to selectively expand only the pages relevant to a query back into their uncompressed form, at configurable 5x, 10x, or 15x compression ratios, letting it reason over long documents while processing far less raw context per query. It is a research model released alongside its arXiv paper and reference code, not a product, and is not described as instruction-tuned for general chat. At 9.4 billion parameters it fits on a single consumer GPU, especially once quantized. Context length is 262,144 tokens. It is released under the Apple Machine Learning Research Model License, a custom license restricted strictly to non-commercial research purposes, with no commercial exploitation permitted. It was published in September 2026.
233 downloads
114 likesSpecifications
- Publisher
- Apple
- Parameters
- 9.4B
- Architecture
- Qwen3_5ForConditionalGeneration
- Context Length
- 262,144 tokens
- Vocabulary Size
- 248,320
- Release Date
- 2026-09-21
- License
- apple-amlr
Get Started
HuggingFace
How Much VRAM Does LensVLM 9B Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| Q2_K | 3.40 | 4.6 GB | 38.7 GB | 4.00 GB | 2-bit quantization with K-quant improvements |
| Q3_K_S | 3.50 | 4.7 GB | 38.8 GB | 4.12 GB | 3-bit small quantization |
| Q3_K_M | 3.90 | 5.2 GB | 39.3 GB | 4.59 GB | 3-bit medium quantization |
| Q4_0 | 4.00 | 5.3 GB | 39.4 GB | 4.70 GB | 4-bit legacy quantization |
| Q4_K_M | 4.80 | 6.2 GB | 40.3 GB | 5.65 GB | 4-bit medium quantization — most popular sweet spot |
| Q5_K_M | 5.70 | 7.3 GB | 41.4 GB | 6.70 GB | 5-bit medium quantization — good quality/size tradeoff |
| Q6_K | 6.60 | 8.3 GB | 42.4 GB | 7.76 GB | 6-bit quantization, very good quality |
| Q8_0 | 8.00 | 10.0 GB | 44.1 GB | 9.41 GB | 8-bit quantization, near-lossless |
Which GPUs Can Run LensVLM 9B?
Q4_K_M · 6.2 GBLensVLM 9B (Q4_K_M) requires 6.2 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 9+ GB is recommended. Using the full 262K context window can add up to 34.1 GB, bringing total usage to 40.3 GB. 52 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 3070 Ti.
Runs great
— Plenty of headroom
Decent
— Enough VRAM, may be tightWhich Devices Can Run LensVLM 9B?
Q4_K_M · 6.2 GB58 devices with unified memory can run LensVLM 9B, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, MacBook Air 13" M3 (8 GB).
Runs great
— Plenty of headroomWhere to Download LensVLM 9B
Community quantizations of this model — GGUF for llama.cpp, Ollama, and LM Studio, plus AWQ/MLX variants where available.
Related Models
Frequently Asked Questions
- How much VRAM does LensVLM 9B need?
LensVLM 9B requires 6.2 GB of VRAM at Q4_K_M, or 19.4 GB at BF16. Full 262K context adds up to 34.1 GB (40.3 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 9.4B × 4.8 bits ÷ 8 = 5.6 GB
KV Cache + Overhead ≈ 0.6 GB (at 2K context + ~0.3 GB framework)
KV Cache + Overhead ≈ 34.7 GB (at full 262K context)
VRAM usage by quantization
Q4_K_M6.2 GBQ4_K_M + full context40.3 GB- What's the best quantization for LensVLM 9B?
For LensVLM 9B, Q4_K_M (6.2 GB) offers the best balance of quality and VRAM usage. Q4_K_L (6.3 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 3.2 GB.
VRAM requirement by quantization
IQ2_XXS3.2 GBQ2_K4.6 GBQ3_K_L5.4 GBQ4_K_M ★6.2 GBQ4_K_L6.3 GBBF1619.4 GB★ Recommended — best balance of quality and VRAM usage.
- Can I run LensVLM 9B on a Mac?
LensVLM 9B requires at least 3.2 GB at IQ2_XXS, which exceeds the unified memory of most consumer Macs. You would need a Mac Studio or Mac Pro with a high-memory configuration.
- Can I run LensVLM 9B locally?
Yes — LensVLM 9B can run locally on consumer hardware. At Q4_K_M quantization it needs 6.2 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is LensVLM 9B?
At Q4_K_M, LensVLM 9B can reach ~773 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~106 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 6.2 × 0.65 = ~837 tok/s
Estimated speed at Q4_K_M (6.2 GB)
~837 tok/s~106 tok/s~837 tok/s~773 tok/sReal-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
- What's the download size of LensVLM 9B?
At Q4_K_M, the download is about 5.65 GB. The full-precision BF16 version is 18.82 GB. The smallest option (IQ2_XXS) is 2.59 GB.
- Which GPUs can run LensVLM 9B?
52 consumer GPUs can run LensVLM 9B at Q4_K_M (6.2 GB). Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6800, AMD Radeon RX 6800 XT, AMD Radeon RX 7600. 40 GPUs have plenty of headroom for comfortable inference.
- Which devices can run LensVLM 9B?
59 devices with unified memory can run LensVLM 9B at Q4_K_M (6.2 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Apple iPhone 17 Pro, Asus ROG Flow Z13 (2025, Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.