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
NVIDIA
Total parameters
3,975,229,440
Exact file size
3,673,288,864 bytes
Quantization / precision
Q5_K_M
Weight repository
nvidia/NVIDIA-Nemotron-Labs-Dory1.0-4B-GGUF
Original model repository
nvidia/NVIDIA-Nemotron-Labs-Dory1.0-4B-BF16
Weight revision
0963c8000791e77b2b1b49aa387d92a6da9988d6
Model revision
b366b87f762ea334ba3eb55912031baba1515810
Runtime notes
Confirm model architecture and quantization support in your runner.
Recorded release date
2026-10-16
Date evidence
NVIDIA model card states 16 October 2026 release. 3,975,229,440 parameters; official complete Q5_K_M GGUF. Recurrent compute is adjustable; scores below use two loops.
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.
GPQA Diamond
72.41 / 100
Hard science questions. This is a science reasoning test, not an overall intelligence score.
Benchmark efficiency: 19.713 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: NVIDIA BF16, two recurrent loops, no tools, NeMo Gym / Nemo Evaluator SDK. Sampling settings and repeat count are not stated.
Publisher result, not an independent evaluation. Weight bytes exclude runtime and KV-cache memory. Maximum context does not establish device compatibility. BF16 scores are not Q5_K_M quality measurements. OpenMDW-1.1. The card requires Dory-enabled llama.cpp PR #30306 or vLLM PR #60149 while upstream support is pending; do not assume stock Ollama or LM Studio support.
Writing code to solve scientific problems. Published results can depend on the test scaffold.
Benchmark efficiency: 9.512 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: NVIDIA BF16, two recurrent loops, NeMo Gym / Nemo Evaluator SDK; eight repeats for SciCode. Sampling details and uncertainty are not stated.
Publisher result, not an independent evaluation. Weight bytes exclude runtime and KV-cache memory. Maximum context does not establish device compatibility. BF16 scores are not Q5_K_M quality measurements. OpenMDW-1.1. The card requires Dory-enabled llama.cpp PR #30306 or vLLM PR #60149 while upstream support is pending; do not assume stock Ollama or LM Studio support.
Agentic terminal-use task success. Publisher-reported percentages are retained, but agent scaffold, tools, timeouts and reasoning budget can materially affect results.
Benchmark efficiency: 4.927 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: NVIDIA BF16, two recurrent loops, Terminus-2 harness. Sampling settings and uncertainty are not stated.
Publisher result, not an independent evaluation. Weight bytes exclude runtime and KV-cache memory. Maximum context does not establish device compatibility. BF16 scores are not Q5_K_M quality measurements. OpenMDW-1.1. The card requires Dory-enabled llama.cpp PR #30306 or vLLM PR #60149 while upstream support is pending; do not assume stock Ollama or LM Studio support.
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 3.673 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 3.673 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.