Acrab Agent Box Mini PC: CATL-Backed Startup Challenges Nvidia DGX Spark

#ai #hardware #startup

Originally published at https://tekmag.thsite.top/acrab-agent-box-mini-pc-catl-backed-startup-challenges-nvidia-dgx-spark/

Singapore-based startup Acrab has unveiled the Agent Box, a mini PC powered by its custom G≡LIX 1 chip that claims to run 100-billion-parameter AI models entirely on-device at one-fifth the cost of Nvidia's DGX Spark. Backed by CATL, the world's largest EV battery maker, the device delivers 700 TOPS of processing power in a compact form factor.

Acrab announced the Agent Box on July 23, 2026, according to PC Watch, which first reported the news. The company positioned the device as a direct competitor to Nvidia's $5,000 DGX Spark, claiming similar performance at roughly one-fifth the price and half the power consumption. TechRadar Pro and Robotics Media covered the announcement in early August 2026.

The startup has reportedly raised over $350 million, per Robotics Media, though no official funding announcement was located. The company's website at acrab.ai confirms the product focus but does not list detailed specifications on its homepage. CATL's involvement appears to be inferred from copyright text visible in the announcement video rather than a formal press statement from the battery manufacturer.

Key Takeaways

Singapore-based Acrab unveiled the Agent Box mini PC on July 23, 2026
Custom G≡LIX 1 chip built on 5nm process with 20-core Arm CPU and 3 TFLOPS GPU
Claims 700 TOPS processing power, comparable to Nvidia DGX Spark at one-fifth the cost
Runs 100B parameter models entirely on-device without cloud reliance
Internal benchmark shows 1,416.8 tokens/sec prefill—7.5x faster than Mac mini M4 Pro
Backed by CATL, the world's largest EV battery manufacturer

The Agent Box hardware

The Agent Box centers on Acrab's proprietary G≡LIX 1 system-on-chip, fabricated on a 5nm process node. The chip integrates a 20-core Arm CPU, a GPU delivering 3 TFLOPS of compute, and a multicore neural processing unit optimized for large language model inference with VAE vector acceleration.

Memory architecture includes 273 GB/s unified memory bandwidth, 8 MB of L1 cache, and 768 GB/s of L2 cache bandwidth. Acrab claims the total processing power reaches 700 TOPS, making it suitable for running large language models entirely on-device without relying on cloud infrastructure.

Performance claims against Nvidia and Apple

Acrab's primary competitive claim targets Nvidia's DGX Spark, a $5,000 AI-focused mini computer. According to the company's internal testing, the Agent Box achieves performance close to the DGX Spark while consuming roughly half the power and costing approximately one-fifth of the price.

In specific benchmarks, Acrab reported a prefill rate of 1,416.8 tokens per second using the Gemma 26B A4B model with a 40,000 token KV cache and 10,000 token input. The company states this is 7.5 times faster than an Apple Mac mini M4 Pro, which achieved 188.9 tokens per second on the same workload. No independent benchmark validation has been found for these claims.

Software stack and use cases

The Agent Box runs a software stack that includes optimized AI runtimes, developer toolchains, OS primitives, and orchestration frameworks designed for autonomous agent use cases. Acrab demonstrated voice-command-driven 3D modeling and printing, robotic vacuum control, and smart home integrations including lights, air conditioning, and smart locks.

Robotics Media reports that Acrab plans to license the platform to manufacturers of AI NAS boxes, AI PCs, smart vehicles, and industrial or service robots. The company has not publicly announced specific pricing or availability dates for the Agent Box.

Who is Acrab and what is their funding situation?

Acrab is a Singapore-based startup that has not publicly disclosed detailed company history or leadership information on its website. Robotics Media reports the company has quietly raised more than $350 million, but no press release or funding announcement was located to verify this figure independently.

The CATL connection derives from footer copyright text mentioning "Contemporary Amperex Technology" on the PC Watch announcement video page. This appears to be an indirect attribution rather than a formal partnership statement from the battery manufacturer, which is the world's largest EV battery producer.

What remains unclear about the Agent Box

Several details about the Agent Box remain unverified or unavailable. No independent benchmark reviews have been published. Pricing and availability have not been clearly stated across any source. The Nvidia DGX Spark comparison is based solely on Acrab's own claims. The extent of CATL's involvement is unclear, and no official statement from the battery maker confirms their role.

FAQ

Q: What is the Acrab Agent Box?

A: The Agent Box is a mini PC unveiled by Singapore-based startup Acrab on July 23, 2026. It is designed to run large language models up to 100 billion parameters entirely on-device.

Q: How does the Agent Box compare to Nvidia's DGX Spark?

A: Acrab claims the Agent Box delivers performance close to the $5,000 DGX Spark at approximately one-fifth the cost and half the power consumption. These claims come from internal testing and have not been independently verified.

Q: What chip powers the Agent Box?

A: The device uses Acrab's proprietary G≡LIX 1 system-on-chip, built on a 5nm process. It features a 20-core Arm CPU, 3 TFLOPS GPU, and multicore NPU for LLM inference, claiming 700 TOPS of total processing power.

Q: Is CATL officially involved with Acrab?

A: The connection appears indirect. Copyright text in the announcement video references "Contemporary Amperex Technology," but no press release or official statement from CATL confirms their involvement.

Q: When will the Agent Box be available and how much will it cost?

A: Pricing and availability have not been announced. Acrab has not released a launch date or retail price for the device.

Conclusion

The Acrab Agent Box represents an ambitious entry into the competitive field of local AI computing. If the company's performance claims hold up under independent testing, the device could make running large language models more accessible outside data centers. The CATL backing and reported $350 million in funding suggest significant resources behind the effort. However, without verified benchmarks, confirmed pricing, or clear availability information, the Agent Box remains a promising concept awaiting real-world validation.