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 (Cerebras) and Tenstorrent Blackhole (Tenstorrent) both sit in Chips & Compute. Cerebras WSE-3 is the cheaper entry point at Free tier / developer tier pricing available (contact/partner integrations). Tenstorrent Blackhole shipped more tracked changes in the last 30 days (17 vs 9). Every value below comes from the latest crawl of the vendors' own pages.
| Attribute | Cerebras WSE-3 Cerebras | Tenstorrent Blackhole Tenstorrent |
|---|---|---|
| Segment | Chips & Compute | Chips & Compute |
| Version | WSE-3 | Blackhole |
| Entry cost | Free tier / developer tier pricing available (contact/partner integrations) | from $999/unit |
| Pricing tiers | Developer Tier free | Cards $999 · Workstations (TT-QuietBox) $9999 · Servers (Galaxy) $70000 |
| 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 | Gemma-2 · Gemma4 · GLM-5.2 · Kimi-K3 · Llama-3.3-70B · Qwen3-32B · Tensix cores · tt-metal / tt-nn |
| Context window | — | — |
| Public API | ||
| Routes models | ||
| Changes / 30d | 9 | 17 |
| 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 | RISC-V cores, open tt-metal stack, Ethernet-scaled cards, IP licensing, Passive, active and liquid cooling options, High-performance AI processing, Liquid-cooled workstations (TT-QuietBox), Scale-out servers (Tenstorrent Galaxy) |
Single-wafer accelerators delivering the fastest published open-model token rates.
Full Cerebras WSE-3 record →Open-source software stack and RISC-V based accelerators sold as cards and IP.
Full Tenstorrent Blackhole record →