vs
Wafer vs Boundless
Wafer and Boundless both tune the serving stack for open models. Wafer uses AI agents that keep re-tuning, and Boundless assigns its own engineers.
By The Subconscious Team · Updated
Wafer vs Boundless: key differences
Both serve GLM-5.3 and DeepSeek V4.1 Flash with up to 1M context, bill serverless use per token from prepaid credits, and offer OpenAI and Anthropic compatible APIs. Wafer's agents test configs across batching, decoding, quantization, kernels and hardware, then keep re-tuning a dedicated deployment on NVIDIA or AMD as load changes. Artificial Analysis measures it at about 176 tokens per second on GLM-5.3 and 152 on DeepSeek V4.1 Flash. Boundless engineers pick a model on your evals, tune precision, batching, caching and speculative decoding on reserved capacity, and run it against uptime, throughput and latency targets in the contract. Vercel AI Gateway measures Boundless at 134 tokens per second on GLM-5.3 and 121 on DeepSeek V4.1 Flash. The two sets of figures come from different benchmarks.
Pick Wafer for an automated tuning loop that adapts as traffic and hardware change, or to run on AMD GPUs. Pick Boundless for GPT and Claude on the same key as open models, a free two-week trial on your own traffic, or targets written into a contract.
What Wafer and Boundless do
Wafer
Wafer builds AI agents that act as GPU performance engineers, then sells inference on the stacks those agents tune. The company came out of Y Combinator's Summer 2025 batch as a "Cursor for CUDA" that turned slow PyTorch into custom kernels. Founders Emilio Andere and Steven Arellano are based in San Francisco. Wafer raised a $4M seed led by Fifty Years in April 2026 and a $40M Series A co-led by Marathon and Chemistry in September 2026 at a $220M post-money valuation, with AMD Ventures and Wing joining. Its agents profile a workload, generate candidate configs across batching, decoding, quantization, engines, kernels and hardware, measure each one and deploy the winner.
Example models: GLM-5.3, DeepSeek V4.1 Flash
Full Wafer profileBoundless
Boundless launched in September 2026 as an inference partner for AI-native startups. It sells two things. The Model API serves pre-tuned models on shared capacity, billed per token from prepaid credits with a budget per API key. Tailored inference reserves capacity for one customer, and Boundless engineers pick the model on the customer’s own evals, tune precision, batching, caching, speculative decoding and parallelism, then run the deployment against uptime, throughput and latency targets written into the contract. A free two-week evaluation tests those targets on real traffic against the customer’s current provider first.
Example models: GLM-5.3, DeepSeek V4.1 Flash, Kimi K3
Full Boundless profileShould you choose Wafer or Boundless?
Wafer vs Boundless at a glance
| Attribute | Wafer | Boundless |
|---|---|---|
| Model access | Open weights | Open weights, plus closed GPT and Claude |
| Flagship models | GLM-5.3, DeepSeek V4.1 Flash | GLM-5.3, DeepSeek V4.1 Flash, Kimi K3 |
| Speed | ~176 tok/s on GLM-5.3; 2–2.8x vs stock vLLM or SGLang (vendor) | 342 tok/s on DeepSeek V4.1 Flash 1017 (Vercel) |
| Price | Per token, prepaid credits | $0.04–$2.30 in, $0.21–$11.40 out per 1M (open) |
| Customization | Agent-tuned dedicated deployments | Serving-stack tuning, no fine-tuning |
| Deployment | Serverless, dedicated | Serverless API, reserved tailored capacity |
| Long context | Up to 1M (GLM-5.3, DeepSeek V4.1 Flash) | 1M context on most models |
Frequently asked questions
What is the difference between Wafer and Boundless?
Wafer and Boundless both tune the serving stack for open models. Wafer uses AI agents that keep re-tuning, and Boundless assigns its own engineers.
When should I choose Wafer over Boundless?
Agent-driven continual re-tuning; Deployments on NVIDIA or AMD; Dedicated endpoints built around an SLO.
When should I choose Boundless over Wafer?
Engineers who tune and run your deployment; GPT and Claude on the same API; Free two-week trial on real traffic.
Is Wafer or Boundless cheaper?
Wafer: Per token, prepaid credits. Boundless: $0.04–$2.30 in, $0.21–$11.40 out per 1M (open). The cheaper choice depends on the model and workload.
Which has more context, Wafer or Boundless?
Wafer: Up to 1M (GLM-5.3, DeepSeek V4.1 Flash). Boundless: 1M context on most models.
Related comparisons
Running long-horizon agents?
If your agents run past 200K tokens, compare both against Subconscious. Our inference stack treats a long-horizon trace as the primary workload, so speed, cost, and accuracy hold up deep into the trace.