Cards
Add to, or config your own rig. Passive, active and liquid cooling options.
Starting at $999
Introducing
Tenstorrent Galaxy™ superclusters
Multi-server configurations for AI workloads – video, code, LLM – run fast.
Near-linear scaling, preserving quality and interactivity.
Compute for every scale.
Workstations
Whisper-quiet, liquid cooled - run models up to 120b parameters from your desk.
Starting at $9,999
Servers
Sovereign, scale-out servers for production AI.
Starting at $70,000
Flexible IP for Specific Workloads
Accelerate your business with in-use, performant, and flexible IP without the vendor lock-in.
TT in the news
Announcements
Newsroom
Stealthium and Tenstorrent Partner to Deliver Runtime Observability for AI Infrastructure
Jul 30, 2026
Newsroom
Announcements
Tenstorrent Sets New Performance Records, Launches TT- Ascalon S, and Expands Across Japan
Jun 30, 2026
TT in the News
Jim Keller: ‘AI Still Obeys the Old Laws of Compute’
Jun 25, 2026
Install TT-Forge (Beta)
Compile models for Tenstorrent hardware
TT-Forge™ is Tenstorrent’s MLIR-based open-source compiler, built on top of our existing AI software stack, and designed to work with PyTorch, JAX, ONNX, and more. TT-Forge™ is now in public beta, and ready for feedback. File a PR or join the conversation on Discord.
Build an Open Future with Us
AI is changing the laws that once governed computing. Our IP is transparent, our architectures are open, and our software is open source so you can edit, select, fork, and own your silicon future.
Join our developer community
Get the latest info, ask questions, review our open-source repos.
Recent contributions from our bounty program
Build an open future with us. Fix bugs, add features, get paid.
Jul 21
Equal-Count Welford Reduction Optimisation
hard
Jul 15
Performance/precision: atan/asin/acos (fp32)
hard
Jul 10
Optimise `deg2rad` / `rad2deg` (fp32/bf16)
medium
Jul 10
Optimise/improve accuracy for pow(x, y) fp32 (non-integer exponent)
hard
Jun 26
Improve tanh accuracy and performance (fp32)
hard
Jun 26
exp: perf regression (WH/BH) and -NaN edge case (WH)
hard