Cerebras WSE-3
Single-wafer accelerators delivering the fastest published open-model token rates.
Full Cerebras WSE-3 record →Head to head
Cerebras WSE-3 vs NVIDIA Blackwell
Cerebras WSE-3 (Cerebras) and NVIDIA Blackwell (NVIDIA) both sit in Chips & Compute. Cerebras WSE-3 shipped more tracked changes in the last 30 days (10 vs 9). Every value below comes from the latest crawl of the vendors' own pages.
| Attribute | Cerebras WSE-3 Cerebras | NVIDIA Blackwell NVIDIA |
|---|---|---|
| Segment | Chips & Compute | Chips & Compute |
| Version | WSE-3 | GB200 NVL72 |
| Entry cost | Free tier · from $10/unit | Pricing not publicly available (enterprise hardware/rack scale) |
| Pricing tiers | Free Trial free · Developer $10 · Enterprise free | — |
| Model stack | Codex-Spark · Gemma · Gemma-4-31B · GLM-4.7 · GPT 5.6 Sol · GPT-5.6-Sol-Ultrafast · Kimi K2.6 · Llama · Llama / Qwen / GPT-OSS · Qwen · WSE-3 | Blackwell · Blackwell B200 · Grace CPU |
| Context window | — | — |
| Public API | ||
| Routes models | ||
| Changes / 30d | 10 | 9 |
| Origin | Global | Global |
| Capabilities | wafer-scale engine, record token/s on open models, CS-3 systems, training clusters, Wafer-scale AI inference, Cerebras PyTorch, ModelZoo, Low-level SDK | 72-GPU NVLink domain, FP4 inference, HBM3E 192GB, CUDA ecosystem, Second-generation Transformer Engine, Fifth-generation NVLink, FP4 AI Support, Tensor Cores with microscaling formats |
NVIDIA Blackwell
NVL72 rack-scale coherent GPU domains, the default substrate for frontier training and inference.
Full NVIDIA Blackwell record →