What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
As of October 7, 2026, Liquid AI has not documented a way to download and run d1 on a consumer laptop, desktop, or phone. Its October 5 announcement describes d1 as available through Liquid AI’s API, with text access also offered through Vercel and OpenRouter at launch. The announcement gives no d1 weights, local runtime, RAM or VRAM minimum, or supported-device list. Liquid AI’s claims about running models on consumer hardware apply to its broader Liquid Foundation Model (LFM) family, not specifically to d1.
What d1 does—and how the announced version is accessed
Liquid AI describes d1 as a decision model: it takes unstructured text, images, or both, plus one or more questions, and returns probabilities for possible answers in a single forward pass. It does not generate answer tokens in the manner of a conventional text-generation model. Its described answer formats include yes/no, choosing among labels, and scoring on a scale. A request can ask several questions about the same input.
As an Amazon Associate I earn from qualifying purchases.
The October 5 announcement says developers can access d1 through the Liquid AI API. It also names Vercel and OpenRouter as text-access routes at launch, with vision support on those services to follow. The API example includes image input as a base64 data URL. These are service routes, not evidence of a local model package.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minuteWhat is—and is not—known about local requirements
The d1 announcement does not specify a downloadable model, local inference runtime, storage requirement, RAM or VRAM minimum, or supported consumer devices. Without those details, there is no evidence-based way to say that a particular laptop, desktop GPU, or phone is sufficient. The release does not document local d1 inference; that is different from proving that no future d1 version could run locally.
#1 Best Overall
- Delivers 25 TFLOPS of AI inference performance using four MemryX MX3 accelerators in a compact M.2 2280 M-key module.
- Connects through PCI Express Gen 3 and above for straightforward integration into embedded and edge systems with an M.2 M-key slot.
- Includes 42 MB of distributed on-chip weight storage to support low-latency, batch-1 pipelined inference.
- Designed for low-power deployment with a 1 W–12 W operating range and host-managed airflow requirements.
- Supports ONNX, TensorFlow, Keras, and TFLite models through the MemryX software stack on Linux and Windows hosts.
Liquid AI makes broader device claims for its Liquid Foundation Models. The company says LFMs are designed for CPU, GPU, and NPU deployment, including phones and laptops; its portfolio descriptions cover variants from hundreds of millions to a few billion parameters. Its pricing page says some general LFM variants are under 1 GB. Those are portfolio-level statements, not d1 specifications, and they do not establish that d1 fits any of those hardware profiles.
Choose between API access and a separate local LFM
| Consideration | d1 through an API | A separate local LFM |
|---|---|---|
| Access and weights | API availability is described in Liquid AI’s October 5, 2026 announcement; a local d1 download is not documented there. | Liquid AI describes local deployment and open-weight availability for its LFM portfolio. Check the exact model’s current download and license details. |
| Hardware and runtime | d1-specific device, memory, and runtime requirements are not stated in the announcement. | Liquid AI describes CPU, GPU, and NPU deployment generally; the particular model’s bundle, runtime, quantization, and device requirements need checking. |
| Input and task | The announcement describes text and image input and probability-based decisions. | Capabilities depend on the selected LFM; the portfolio-level device claims do not establish that a particular LFM is a d1 substitute for every decision task. |
| Cost and resources | The announced API billing is based on input tokens; image inputs also count toward that total. | Local use shifts the practical question to the selected model’s hardware and runtime needs; a d1 API price is not a measure of local inference cost. |
| Offline use | The announcement establishes API access, not offline operation. | Local deployment may support offline use for a compatible model, but confirm the selected model and runtime rather than assuming d1 compatibility. |
Liquid AI’s FAQ describes the LFM portfolio and deployment stack at a general level. Its LFM Cloud Demos concern separate device, edge-GPU, and private-cloud demonstrations; they are not d1 hardware tests. If you need a Liquid model that runs locally now, choose a specific downloadable LFM and verify its own package, runtime, quantization, and task fit.
Rank #2
- ✅Powered by 26 Tera-Operations Per Second (TOPS) Hailo-8 AI Processor. 2.5W typical power consumption
- ✅Scalable, enabling simultaneous processing of multi-streams & multi-models
- ✅Enabling real-time, low latency and high-efficiency AI inferencing on the edge devices
- ✅Supports TensorFlow, TensorFlow Lite, ONNX, Keras, Pytorch frameworks
- ✅Supports Linux and Windows. Supports the temperature range of -40°C to 85°C
API latency and image billing are not hardware requirements
Liquid AI reports 200–300 ms for a text decision in its announcement. That is a company-reported API latency, not a measurement of inference on a consumer device. The post also describes comparisons across six applications, with each application run once per model on October 5, 2026; those vendor-reported results do not establish local hardware performance.
For d1 API requests, Liquid AI says billing is based on input tokens, with no output-token charge. Its image accounting is 1.5 tokens per 32×32-pixel patch; the announcement gives 1,536 tokens for a 1024×1024 image. Each question is billed as its own prompt, including its text and all images. This is the API cost model, not a description of local inference economics.
Rank #3
- NVIDIA Volta GV100 Architecture — 4,608 CUDA Cores, 640 1st-Gen Tensor Cores delivering 14 TFLOPS FP32 and 112 TFLOPS deep learning performance for AI training, inference, HPC, and scientific computing workloads
- 32GB HBM2 ECC Memory — 900 GB/s Bandwidth — High-bandwidth memory on a 4096-bit bus with ECC error correction provides the memory capacity and throughput required for the largest AI models, simulations, and datasets
- PCIe 3.0 x16 Interface — 250W TDP — Standard PCIe Gen3 connectivity with passive cooling designed for enterprise rack server deployment in HPE ProLiant, Dell PowerEdge, and Supermicro platforms with adequate chassis airflow
- NVLink — Scale to 96GB Unified Memory — Connect two V100 GPUs via NVLink at 300 GB/s bi-directional bandwidth to scale GPU memory from 32GB to 96GB for larger AI training and HPC workloads
- Multi-Precision Computing — Supports FP64 (7 TFLOPS), FP32 (14 TFLOPS), FP16 (112 TFLOPS) and INT8 precision modes for flexible deployment across training, inference, and scientific simulation workloads
What to watch for if local d1 is essential
Liquid AI says it plans to release open weights for upcoming models, but the October 5 announcement gives no release date and does not promise that a named d1 version is included. Until a d1-specific release documents downloadable weights and local compatibility, treat d1 as an API-accessible service. Reassess only when Liquid AI publishes the exact model package and requirements.
Quick Recap
Rank #4
- Powered by 26 Tera-Operations Per Second (TOPS) Hailo-8 AI Processor.
- 2.5W typical power consumption
- Enabling real-time low latency and high-efficiency AI inferencing on the edge devices
- Supports TensorFlow TensorFlow Lite, ONNX, Keras, Pytorch frameworks
- Supports Linux and Windows.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
