Unverified50% confidenceFactExact time
In local AI model tests, the 64GB RTX Spark test machine could not load large models in the 120B to 125B+ range, such as DeepSeek-V3 and Qwen 3.8 Flash Next.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/17/2026
First Seen
Valid until: 1/15/2027
Sources
Related Entities
Related Claims
UnverifiedThe RTX Spark's 128 GB of unified memory can accommodate 70B-level models in terms of capacity (about 35-45 GB after 4-bit quantization), but whether it can achieve smoothness close to a cloud Agent remains to be verified by actual evaluation.77% similarUnverifiedThe UP host's test environment consisted of an RTX 5070 GPU (12GB VRAM), the Qwen3 series 9B parameter model, and a 64K context length.73% similarUnverifiedThe RTX Spark can be equipped with up to 20 CPU cores and 128GB of LPDDR5X memory, enough to run large models and even agent workloads locally on thin and light devices.71% similarUnverifiedRunning the 125B-parameter Flash Next model on an RTX 3090 (24GB VRAM) with 24GB RAM, generation speed is stable at 9 to 10 tokens per second, with a peak of 18 for short texts.68% similarUnverifiedUnoptimized baseline token decoding speeds for running Qwen 3 (27B): RTX 5090 ~75, RTX 4090 ~46, RTX 3090 ~40, M5 Max 31.4, DGX Spark 12–17, Ryzen AI Max 11.5 tokens/s.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1055112API
curl https://kongchang.com/api/v1/knowledge/claims/1055112MCP
get_claim(id=1055112)