Cerebras Systems announced it is now running Kimi K2.6, a trillion-parameter AI model, at nearly 1,000 tokens per second, a speed no GPU-based provider has matched. The achievement, independently verified by benchmarking firm Artificial Analysis, clocked in at 981 output tokens per second, making Cerebras 6.7 times faster than the next-fastest GPU-based cloud provider and 23 times faster than the median. For a standard coding request involving 10,000 input tokens, Cerebras delivered the full response in 5.6 seconds, compared to 163.7 seconds on the official Kimi endpoint, a 29-fold improvement in time to final answer.

The announcement comes less than a week after Cerebras completed the largest tech IPO of 2026, giving the Sunnyvale-based chipmaker a $95 billion market cap and $5.55 billion in IPO proceeds to fuel its expansion. This milestone marks a critical inflection point for the company, which has long battled the perception that its unorthodox wafer-scale chips, while blindingly fast, could only handle small and mid-sized models. Kimi K2.6 is the first trillion-parameter open-weight model Cerebras has ever served in production.

Why Is Cerebras Choosing a Chinese-Built Model as Its Flagship?

Kimi K2.6, released on April 20 by Beijing-based Moonshot AI, is a Mixture-of-Experts model that has rapidly established itself as the most capable open-weight model available for coding and agentic tasks. The model tops SWE-Bench Pro at 58.6, outperforming Claude Opus 4.6 and matching GPT-5.4, while posting leading scores on agentic benchmarks like Humanity's Last Exam and DeepSearchQA. Its architecture uses 32 billion activated parameters per token out of a total of 1 trillion, with 384 experts, of which 8 are selected plus 1 shared per forward pass, operating over a 256,000-token context window.

In practical terms, K2.6 is one of the first open-weight models that enterprises can plausibly use as a drop-in replacement for expensive, capacity-constrained closed-source APIs from Anthropic and OpenAI, particularly for coding and agentic workloads. The version 2.6 release extends K2.6's capabilities from front-end design into full-stack workflows, including authentication, database operations, and long-horizon agent execution.

"They're very motivated, first of all, to have an alternative to Anthropic," said James Wang, director of product marketing at Cerebras. "Anthropic's models are fantastic. I use them. I'm sure you probably use them. But they're quite expensive, and they're constantly running out of capacity."

James Wang, Director of Product Marketing at Cerebras Systems

The geopolitical dimension of this arrangement is worth noting. Kimi K2.6 is a Chinese-developed model being served by an American chipmaker to American enterprise customers. Moonshot AI operates out of Beijing, and K2.6's adoption in the West arrives during a period of heightened scrutiny of Chinese AI companies in the U.S. market. Enterprise buyers with strict compliance requirements, particularly those in financial services, healthcare, and defense, will need to evaluate this dimension alongside the model's technical capabilities.

How Do Wafer-Scale Chips Solve the Speed Problem That GPUs Cannot?

Understanding why Cerebras achieves these speeds requires understanding what makes its hardware fundamentally different from anything else on the market. Most AI inference today runs on clusters of Nvidia GPUs, typically organized in racks of 72 GPUs, what Nvidia markets as the NVL72 configuration. In these setups, the model's parameters are distributed across many discrete chips connected by high-speed networking fabric. Data must constantly shuttle between chips, and the interconnect bandwidth between GPUs becomes a bottleneck, particularly for large models with hundreds of billions or trillions of parameters.

Cerebras takes a radically different approach. Its Wafer-Scale Engine 3 is a single chip the size of an entire silicon wafer, roughly the size of a dinner plate, containing 44 gigabytes of on-chip SRAM. Unlike the high-bandwidth memory used in GPUs, SRAM sits directly on the processor die, offering dramatically lower latency and higher bandwidth for data access. For Kimi K2.6, Cerebras stores the model's weights in their original 4-bit precision while performing computation at 16-bit floating point.

The weights are distributed across multiple wafers in a cluster of approximately 20 CS-3 systems, with activations streamed between them. Critically, all the experts for a given MoE (Mixture-of-Experts) layer are placed on the same wafer, meaning the all-to-all communication required for expert routing happens at SRAM speeds. According to Cerebras' technical description, the on-wafer network fabric delivers over 200 times the bandwidth of NVLink on NVL72.

"Our single units are much larger and much higher capacity, they're on the order of 20 racks, as opposed to 72 GPUs," explained James Wang. "Each layer in the transformer can, in effect, serve a separate user simultaneously. They're just like a queue, like you're queuing for bagels or something, they're all occupying a different part of the hardware. But because they move across so fast, the actual experience, tokens per second, single user, on your end is still what you're used to."

James Wang, Director of Product Marketing at Cerebras Systems

Combined with custom kernels and speculative decoding, this architecture allows Cerebras to serve the trillion-parameter MoE model at close to 1,000 tokens per second, a speed the company calls a world record achievable only with wafer-scale hardware.

How to Evaluate Cerebras' Inference Offering for Your Enterprise

• Performance Benchmarks: Compare token-per-second throughput and time-to-final-answer metrics against your current provider. Cerebras' 981 tokens per second and 5.6-second response times for standard coding requests represent significant improvements over GPU-based alternatives.
• Model Compatibility: Assess whether open-weight models like Kimi K2.6 meet your compliance and capability requirements. K2.6 supports full-stack workflows including authentication, database operations, and long-horizon agent execution, making it suitable for complex enterprise tasks.
• Compliance and Data Residency: Evaluate the geopolitical and regulatory implications of using a Chinese-developed model served through American infrastructure, particularly if your organization operates in financial services, healthcare, or defense sectors.
• Cost and Capacity Reliability: Consider whether the speed improvements justify switching from established providers like Anthropic, especially if your current provider frequently runs out of capacity during peak usage periods.

Who Is Already Using Cerebras' Trillion-Parameter Inference?

Cerebras is not opening K2.6 to the general public. Instead, the company is positioning this as an enterprise-first offering, with Fortune 500 companies in software, financial services, and healthcare currently running cloud trials of their production workloads on the platform. The enterprise-first approach is deliberate. Cerebras has historically prioritized its largest customers over consumer-facing APIs, in part because of hardware capacity constraints.

"These are logos that you've definitely heard of," said James Wang, though he declined to identify specific customers due to confidentiality agreements.

James Wang, Director of Product Marketing at Cerebras Systems

The announcement signals to Wall Street that Cerebras intends to compete not just at the frontier of speed, but at the frontier of model scale. With a freshly minted $95 billion market cap and $5.55 billion in IPO proceeds, the company has the capital to expand its wafer-scale manufacturing and compete directly with GPU-based cloud providers for enterprise inference workloads. The achievement demonstrates that specialized hardware designed specifically for inference can deliver performance gains that general-purpose GPUs simply cannot match, even at trillion-parameter scale.