[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-10911":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":9,"totalLinesOfCode":9,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":9,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":9,"rankLanguage":9,"license":9,"archived":22,"fork":22,"defaultBranch":23,"hasWiki":22,"hasPages":22,"topics":24,"createdAt":9,"pushedAt":9,"updatedAt":38,"readmeContent":39,"aiSummary":40,"trendingCount":16,"starSnapshotCount":16,"syncStatus":41,"lastSyncTime":42,"discoverSource":43},10911,"tt-metal","tenstorrent\u002Ftt-metal","tenstorrent",":metal: TT-NN operator library, and TT-Metalium low level kernel programming model.",null,"https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal","C++",1509,483,20,3843,0,14,30,62,42,21.05,false,"main",[25,26,27,28,7,29,30,31,32,33,34,35,36,37],"llama","llm","metal","stable-diffusion","accelerator","scale-out","kernels","ai","deepseek","gpu","img-gen","video-gen","cuda","2026-06-12 02:02:28","[![tt-metal CI](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Factions\u002Fworkflows\u002Fsanity-tests.yaml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Factions\u002Fworkflows\u002Fsanity-tests.yaml)\n[![Ask DeepWiki](https:\u002F\u002Fdeepwiki.com\u002Fbadge.svg)](https:\u002F\u002Fdeepwiki.com\u002Ftenstorrent\u002Ftt-metal)\n\n\u003Cdiv align=\"center\">\n\n\u003Ch1>\n\n[Hardware](https:\u002F\u002Ftenstorrent.com\u002Fhardware\u002Fblackhole) | [Install](.\u002FINSTALLING.md) |  [Discord](https:\u002F\u002Fdiscord.gg\u002FtvhGzHQwaj) | [Join Us](https:\u002F\u002Fboards.greenhouse.io\u002Ftenstorrent\u002Fjobs\u002F4155609007) | [Bounty $](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Fissues?q=is%3Aissue%20state%3Aopen%20label%3Abounty)\n\n\u003C\u002Fh1>\n\n\u003Cimg src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Ftenstorrent\u002Ftt-metal\u002Fmain\u002Fdocs\u002Fsource\u002Fcommon\u002F_static\u002Ftt_nn_w_logo.png\" alt=\"ttnn logo\" height=\"180\"\u002F>\n\n**TT-NN** is a Python & C++ Neural Network OP library.\n\n\u003Ch3>\n\n[API Reference](https:\u002F\u002Fdocs.tenstorrent.com\u002Ftt-metal\u002Flatest\u002Fttnn\u002Findex.html) | [Model Demos](.\u002Fmodels\u002Fdemos\u002F)\n\n\u003C\u002Fh3>\n\n\u003C\u002Fdiv>\n\n## Featured Models\n\nThe Models team is focused on developing the following models, optimizing them for performance, accuracy, and compatibility. Follow each model link for more details.\n\n>[!IMPORTANT]\n> For a **full model list** see the **[Model Matrix](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Fblob\u002Fmain\u002Fmodels\u002FREADME.md)**, or visit the **[Developer Hub](https:\u002F\u002Ftenstorrent.com\u002Fdevelopers)**.\n\n>[!NOTE]\n> Performance Metrics:\n> - Time to First Token (TTFT) measures the time (in milliseconds) it takes to generate the first output token after input is received.\n> - T\u002FS\u002FU (Tokens per Second per User): Represents the throughput of first-token generation after prefill. It is calculated as 1 \u002F inter-token latency.\n> - T\u002FS (Tokens per Second): Represents total token throughput, calculated as T\u002FS = T\u002FS\u002FU x batch size.\n> - TP (Tensor Parallel) and DP (Data Parallel): Indicate the parallelization factors across multiple devices.\n> - Reported LLM Performance: Based on an input sequence length of 128 tokens for all models.\n> - Performance Data Source: Metrics were collected using the tt-metal model demos (linked above). Results may vary when using other runtimes such as the vLLM inference server.\n\n### [Llama 3.3 70B (TP=32)](.\u002Fmodels\u002Fdemos\u002Fllama3_70b_galaxy)\n| Batch | Hardware | TTFT (MS) | T\u002FS\u002FU | Target\u003Cbr>T\u002FS\u002FU | T\u002FS | TT-Metalium Release | vLLM Tenstorrent Repo Release |\n|-------|----------|-----------|-------|-----------------|-----|---------------------|-------------------------------|\n| 32    | [Galaxy (Wormhole)](https:\u002F\u002Ftenstorrent.com\u002Fhardware\u002Fgalaxy) | 53      | 72.5  | 80              | 2268.8  | [v0.65.0-rc7](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Ftree\u002Fv0.65.0-rc7) | [59be953](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Fvllm\u002Ftree\u002F59be953f2bbd21e227f9ef4b779f545f9c3bf599\u002Ftt_metal) |\n\n### [Qwen 2.5 7B (TP=2)](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Ftree\u002Fmain\u002Fmodels\u002Ftt_transformers)\n| Batch | Hardware | TTFT (MS) | T\u002FS\u002FU | Target\u003Cbr>T\u002FS\u002FU | T\u002FS  | TT-Metalium Release | vLLM Tenstorrent Repo Release |\n|-------|----------|-----------|-------|-----------------|------|---------------------|-------------------------------|\n| 32 | [n300 (Wormhole)](https:\u002F\u002Ftenstorrent.com\u002Fhardware\u002Fwormhole) | 109 | 22.1 | 30 | 707.2 | [v0.62.0-rc35](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Ftree\u002Fv0.62.0-rc35) | [ced0161](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Fvllm\u002Ftree\u002Fced0161dc223e6d8aca5f44a6c43d13070c3fba6\u002Ftt_metal) |\n\n### [Qwen 2.5 72B (TP=8)](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Ftree\u002Fmain\u002Fmodels\u002Ftt_transformers)\n| Batch | Hardware | TTFT (MS) | T\u002FS\u002FU | Target\u003Cbr>T\u002FS\u002FU | T\u002FS | TT-Metalium Release | vLLM Tenstorrent Repo Release |\n|-------|----------|-----------|-------|-----------------|-----|---------------------|-------------------------------|\n| 32 | [QuietBox (Wormhole)](https:\u002F\u002Ftenstorrent.com\u002Fhardware\u002Ftt-quietbox) | 223 | 15.4 | 20 | 492.8 | [v0.62.0-rc25](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Ftree\u002Fv0.62.0-rc25) | [e7c329b](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Fvllm\u002Ftree\u002Fe7c329b1664f8591ae8b4269bed9690726e52a24\u002Ftt_metal) |\n\n### [Whisper (distil-large-v3)](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Ftree\u002Fmain\u002Fmodels\u002Fdemos\u002Faudio\u002Fwhisper)\n| Batch | Hardware | TTFT (MS) | T\u002FS\u002FU | Target\u003Cbr>T\u002FS\u002FU | T\u002FS | TT-Metalium Release |\n|-------|----------|-----------|-------|-----------------|-----|---------------------|\n| 1     | [n150 (Wormhole)](https:\u002F\u002Ftenstorrent.com\u002Fhardware\u002Fwormhole)        | 163       | 105.0  | 45           | 105.0   | [v0.65.0-dev20251208](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Ftree\u002Fv0.65.0-dev20251208) |\n| 1     | [p150 (Blackhole)](https:\u002F\u002Ftenstorrent.com\u002Fhardware\u002Fblackhole)        | 63       | 263.4  |            | 263.4   | [v0.65.0-dev20251208](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Ftree\u002Fv0.65.0-dev20251208) |\n\n### [Mixtral 8x7B (TP=8)](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Ftree\u002Fmain\u002Fmodels\u002Ftt_transformers)\n| Batch | Hardware | TTFT (MS) | T\u002FS\u002FU | Target\u003Cbr>T\u002FS\u002FU | T\u002FS | TT-Metalium Release |\n|-------|----------|-----------|-------|-----------------|-----|---------------------|\n| 32 | [QuietBox (Wormhole)](https:\u002F\u002Ftenstorrent.com\u002Fhardware\u002Ftt-quietbox) | 122 | 24.9 | 33 | 796.8 | [v0.62.0-dev20251015](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Ftree\u002Fv0.62.0-dev20251015) |\n\nBlackhole software optimization is under active development.  Please join us in shaping the future of open source AI! \u003Cbr> [\\[Discord\\]](https:\u002F\u002Fdiscord.gg\u002Ftenstorrent) [\\[Developer Hub\\]](https:\u002F\u002Ftenstorrent.com\u002Fdevelopers)\n\nFor more information regarding vLLM installation and environment creation visit the [Tenstorrent vLLM repository](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Fvllm\u002Fblob\u002Fdev\u002Ftt_metal\u002FREADME.md).\n\n## Model Updates\n\nFor the latest model updates and features, please see [MODEL_UPDATES.md](models\u002Fdocs\u002FMODEL_UPDATES.md)\n\n## Model Bring-Up and Testing\n\nFor information on initial model procedures, please see [Model Bring-Up and Testing](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Fblob\u002Fmain\u002Fmodels\u002Fdocs\u002Fmodel_bring_up.md)\n\n## TT-NN Tech Reports\n\n- [Advanced Performance Optimizations for Models](.\u002Ftech_reports\u002FAdvancedPerformanceOptimizationsForModels\u002FAdvancedPerformanceOptimizationsForModels.md) (updated March 4th, 2025)\n- [ViT Implementation in TT-NN on GS](.\u002Ftech_reports\u002FViT-TTNN\u002Fvit.md)  (updated Sept 22nd, 2024)\n- [LLMs Bring up in TT-NN](.\u002Ftech_reports\u002FLLMs\u002Fllms.md)  (updated Oct 29th, 2024)\n- [CNN Bring up & Optimization in TT-NN](.\u002Ftech_reports\u002FCNNs\u002Fcnn_optimizations.md) (updated Jan 22nd, 2025)\n\n## Benchmarks\n\n- [Matrix Multiply FLOPS on Wormhole and Blackhole](.\u002Ftech_reports\u002FGEMM_FLOPS\u002FGEMM_FLOPS.md)  (updated June 17th, 2025)\n\n---\n\n\u003Cdiv align=\"center\">\n\n\u003Cimg src=\"https:\u002F\u002Fraw.githubusercontent.com\u002Ftenstorrent\u002Ftt-metal\u002Fmain\u002Fdocs\u002Fsource\u002Fcommon\u002Fimages\u002Ftt_refresh_metalium_w_icon.png\" alt=\"TT-Metalium logo\" height=\"180\"\u002F>\n\n**TT-Metalium** is our low-level programming model, enabling kernel development for Tenstorrent hardware.\n\n\u003Ch3>\n\n[Programming Guide](.\u002FMETALIUM_GUIDE.md) | [API Reference](https:\u002F\u002Fdocs.tenstorrent.com\u002Ftt-metal\u002Flatest\u002Ftt-metalium\u002Ftt_metal\u002Fapis\u002Findex.html)\n\n\u003C\u002Fh3>\n\u003C\u002Fdiv>\n\n## Getting started\n\nGet started with [simple kernels](https:\u002F\u002Fdocs.tenstorrent.com\u002Ftt-metal\u002Flatest\u002Ftt-metalium\u002Ftt_metal\u002Fexamples\u002Findex.html).\n\n## TT-Metalium Tech Reports\n\n- [Matrix Engine](.\u002Ftech_reports\u002Fmatrix_engine\u002Fmatrix_engine.md) (updated Sept 6th, 2024)\n- [Data Formats](.\u002Ftech_reports\u002Fdata_formats\u002Fdata_formats.md) (updated Sept 7th, 2024)\n- [Reconfiguring Data Formats](.\u002Ftech_reports\u002Fdata_formats\u002Freconfig_data_format.md) (updated Oct 17th, 2024)\n- [Handling special floating-point numbers](.\u002Ftech_reports\u002FHandling_Special_Value\u002Fspecial_values.md) (updated Oct 5th, 2024)\n- [Allocator](.\u002Ftech_reports\u002Fmemory\u002Fallocator.md) (Updated Dec 19th, 2024)\n- [Tensor Layouts](.\u002Ftech_reports\u002Ftensor_layouts\u002Ftensor_layouts.md) (updated Sept 6th, 2024)\n- [Saturating DRAM Bandwidth](.\u002Ftech_reports\u002FSaturating_DRAM_bandwidth\u002FSaturating_DRAM_bandwidth.md) (updated Sept 6th, 2024)\n- [Flash Attention on Wormhole](.\u002Ftech_reports\u002FFlashAttention\u002FFlashAttention.md) (updated Sept 6th, 2024)\n- [CNNs on TT Architectures](.\u002Ftech_reports\u002FCNNs\u002Fttcnn.md) (updated Sept 6th, 2024)\n- [Ethernet and Multichip Basics](.\u002Ftech_reports\u002FEthernetMultichip\u002FBasicEthernetGuide.md) (Updated Sept 20th, 2024)\n- [Blackhole Bring-Up Programming Guide](.\u002Ftech_reports\u002FBlackhole\u002FBlackholeBringUpProgrammingGuide.md) (Updated Dec 18th, 2024)\n- [Sub-Devices](.\u002Ftech_reports\u002FSubDevices\u002FSubDevices.md) (Updated Jan 7th, 2025)\n\n## Scaleout Tech Reports\n\n- [Programming Mesh of Devices (Scale-Up)](.\u002Ftech_reports\u002FProgramming_Mesh_of_Devices\u002FProgramming_Mesh_of_Devices_with_TT-NN.md) (updated Jan 6th, 2026)\n- [Programming Multiple Meshes (Scale-Out)](.\u002Ftech_reports\u002FProgramming_Multiple_Meshes\u002FProgramming_Multiple_Meshes.md) (updated Jan 19th, 2026)\n- [TT-Fabric Architecture](.\u002Ftech_reports\u002FTT-Fabric\u002FTT-Fabric-Architecture.md) (updated Dec 1st, 2025)\n- [TT-Distributed Architecture](.\u002Ftech_reports\u002FTT-Distributed\u002FTT-Distributed-Architecture-1219.md) (updated Oct 20th, 2025)\n\n## TT-Metalium Programming Examples\n\n### Hello World\n\n- [Hello World! Compute Kernel](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Fblob\u002Fmain\u002Ftt_metal\u002Fprogramming_examples\u002Fhello_world_compute_kernel\u002Fhello_world_compute.md)\n- [Hello World! Data Movement Kernel](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Fblob\u002Fmain\u002Ftt_metal\u002Fprogramming_examples\u002Fhello_world_datamovement_kernel\u002Fhello_world_data_movement.md)\n\n### Add Integers\n\n- [Add 2 Integers in Baby RiscV](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Fblob\u002Fmain\u002Ftt_metal\u002Fprogramming_examples\u002Fadd_2_integers_in_riscv\u002Fadd_2_integers_in_riscv.md)\n- [Add 2 Integers in Compute Kernel](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Fblob\u002Fmain\u002Ftt_metal\u002Fprogramming_examples\u002Fadd_2_integers_in_compute\u002Fadd_2_integers_in_compute.md)\n\n### Simple Tensor Manipulation\n\n- [Sharding](.\u002Ftech_reports\u002Fprog_examples\u002Fshard_data_rm\u002Fshard_data_rm.md)\n- [Padding](.\u002Ftech_reports\u002Fprog_examples\u002Fpad_multi_core\u002Fpad_multi_core.md)\n\n### DRAM Data Movement\n\n- [Dram Loopback Data Movement](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Fblob\u002Fmain\u002Ftt_metal\u002Fprogramming_examples\u002Floopback\u002Fdram_loopback.md)\n\n### Eltwise\n\n- [Eltwise Unary OP in Vector Engine (SFPU)](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Fblob\u002Fmain\u002Ftt_metal\u002Fprogramming_examples\u002Feltwise_sfpu\u002Feltwise_sfpu.md)\n- [Eltwise Binary OP in Matrix Engine (FPU)](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Fblob\u002Fmain\u002Ftt_metal\u002Fprogramming_examples\u002Feltwise_binary\u002Feltwise_binary.md)\n\n### Matmul\n\n- [Matmul OP on a Single_core](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Fblob\u002Fmain\u002Ftt_metal\u002Fprogramming_examples\u002Fmatmul\u002Fmatmul_single_core\u002Fmatmul_single_core.md)\n- [Matmul OP on Multi_core (Basic)](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Fblob\u002Fmain\u002Ftt_metal\u002Fprogramming_examples\u002Fmatmul\u002Fmatmul_multi_core\u002Fmatmul_multi_core.md)\n- [Matmul Multi_core Reuse (Optimized)](.\u002Ftech_reports\u002Fprog_examples\u002Fmatmul_multi_core_optimized\u002Fdata_reuse.md)\n- [Matmul Multi_core Multi-Cast (Optimized)](.\u002Ftech_reports\u002Fprog_examples\u002Fmatmul_multi_core_optimized\u002Fdata_mcast.md)\n\n### Tools and Instruments\n\n#### [TT-NN Visualizer](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Fttnn-visualizer)\nA comprehensive tool for visualizing and analyzing model execution, offering interactive graphs, memory plots, tensor details, buffer overviews, operation flow graphs, and multi-instance support with file or SSH-based report loading.\n\n#### [TT-Exalens](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-exalens)\nThe TT-Exalens repository describes TT-Lensium, a low-level debugging tool for Tenstorrent hardware. It allows developers to access and communicate with Wormhole and Blackhole devices.\n\n#### [TT-SMI](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-smi)\nThe TT-SMI repository describes the Tenstorrent System Management Interface. This command line utility can interact with Tenstorrent devices on host. TT-SMI provides an easy to use interface displaying device, telemetry, and firmware information.\n\n#### [Model Explorer](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Fmodel-explorer)\nThe Model Explorer is an intuitive and hierarchical visualization tool using model graphs. It organizes model operations into nested layers and provides features for model exploration and debugging.\n\n#### [Tracy Profiler](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftracy)\nThe Tracy Profiler is a real-time nanosecond resolution, remote telemetry, hybrid frame, and sampling tool. Tracy supports profiling CPU, GPU, memory allocation, locks, context switches, and more.\n\n#### [Kernel Print Debug](https:\u002F\u002Fdocs.tenstorrent.com\u002Ftt-metal\u002Flatest\u002Ftt-metalium\u002Ftools\u002Fkernel_print.html)\nDPRINT can print variables, addresses, and circular buffer data from kernels to the host terminal or log file. This feature is useful for debugging issues with kernels.\n\n#### [Watcher](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Fblob\u002Fmain\u002Fdocs\u002Fsource\u002Ftt-metalium\u002Ftools\u002Fwatcher.rst)\nWatcher monitors firmware and kernels for common programming errors, and overall device status. If an error or hang occurs, Watcher displays log data of that occurrence.\n\n#### [Inspector](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Fblob\u002Fmain\u002Fdocs\u002Fsource\u002Ftt-metalium\u002Ftools\u002Finspector.rst)\nInspector provides insights into host runtime. It logs necessary data for investigation and allows queries to host runtime data.\n\n## Related Tenstorrent Projects\n- [TT-Forge](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-forge\u002Ftree\u002Fmain)\n- [TT-Forge-FE](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-forge-fe)\n- [TT-Torch](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-torch)\n- [TT-XLA](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-xla)\n- [TT-MLIR](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-mlir)\n- [TT-TVM](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-tvm)\n\n## Latest Releases\n\n| Release | Release Date | FW Version | KMD Version | SMI Version |\n|:---------:|:--------------:|:------------:|:--------:|:--------:|\n| 0.68.0 | ETA Apr 30, 2026 | 19.2.0 | 2.5.0 | 3.0.38 |\n| [0.67.4](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Freleases\u002Ftag\u002Fv0.67.4) | Mar 30, 2026 | 19.2.0 | 2.5.0 | 3.0.38 |\n| [0.67.0](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Freleases\u002Ftag\u002Fv0.67.0) | Mar 25, 2026 | 19.2.0 | 2.5.0 | 3.0.38 |\n| [0.66.0](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Freleases\u002Ftag\u002Fv0.66.0) | Feb 18, 2026 | 19.2.0 | 2.5.0 | 3.0.38 |\n| [0.65.1](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Freleases\u002Ftag\u002Fv0.65.1) | Jan 12, 2026 | 19.2.0 | 2.5.0 | 3.0.38 |\n| [0.65.0](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Freleases\u002Ftag\u002Fv0.65.0) | Dec 15, 2025 | 19.2.0 | 2.5.0 | 3.0.38 |\n| [0.64.5](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Freleases\u002Ftag\u002Fv0.64.5) | Dec  1, 2025 | 18.12.0 | 2.4.1 | 3.0.32 |\n| [0.64.4](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Freleases\u002Ftag\u002Fv0.64.4) | Nov 24, 2025 | 18.12.0 | 2.4.1 | 3.0.32 |\n| [0.64.3](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Freleases\u002Ftag\u002Fv0.64.3) | Nov 14, 2025 | 18.12.0 | 2.4.1 | 3.0.32 |\n| [0.64.0](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Freleases\u002Ftag\u002Fv0.64.0) | Oct 29, 2025 | 18.12.0 | 2.4.1 | 3.0.32 |\n| [0.63.0](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Freleases\u002Ftag\u002Fv0.63.0) | Sep 22, 2025 | 18.8.0 | 2.3.0 | 3.0.28 |\n| [0.62.2](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Freleases\u002Ftag\u002Fv0.62.2) | Aug 20, 2025 | 18.6.0 | 2.0.0 | 3.0.20 |\n| 0.61.0 | Skipped | - | - | - |\n| [0.60.1](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Freleases\u002Ftag\u002Fv0.60.1) | Jul 22, 2025 | 18.6.0 | 2.0.0 | 3.0.20 |\n| [0.59.0](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Freleases\u002Ftag\u002Fv0.59.0) | Jun 18, 2025 | - | - | - |\n| [0.58.0](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Freleases\u002Ftag\u002Fv0.58.0) | May 13, 2025 | - | - | - |\n| [0.57.0](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Freleases\u002Ftag\u002Fv0.57.0) | Apr 15, 2025 | - | - | - |\n| [0.56.0](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Freleases\u002Ftag\u002Fv0.56.0) | Mar 7, 2025  | - | - | - |\n\nVisit the [releases](https:\u002F\u002Fgithub.com\u002Ftenstorrent\u002Ftt-metal\u002Ftree\u002Fmain\u002Freleases) folder for details on releases, release notes, and estimated release dates.\n\n## Tenstorrent Bounty Program Terms and Conditions\nThis repo is a part of Tenstorrent’s bounty program. If you are interested in helping to improve tt-metal, please make sure to read the [Tenstorrent Bounty Program Terms and Conditions](https:\u002F\u002Fdocs.tenstorrent.com\u002Fbounty_terms.html) before heading to the issues tab. Look for the issues that are tagged with both “bounty” and difficulty level!\n\n## License\nTT-Metalium and TTNN are licensed under the Apache 2.0 License, as detailed in [LICENSE](LICENSE) and [LICENSE_understanding.txt](LICENSE_understanding.txt).\n\nSome distributable forms of this project—such as manylinux-compliant wheels—may need to bundle additional libraries beyond the standard Linux system libraries. For example:\n\n- libnuma\n- libhwloc\n- openmpi (when built with multihost support)\n- libevent (when built with multihost support)\n\nThese libraries are bound by their own license terms.\n","tenstorrent\u002Ftt-metal 是一个用于神经网络操作的库，同时提供了一个名为TT-Metalium的底层内核编程模型。该项目采用C++开发，旨在通过优化性能、准确性和兼容性来支持包括Llama和Qwen在内的多种大语言模型。它特别强调了在Tenstorrent硬件上的高效执行，能够显著提升模型的推理速度与吞吐量。适合于需要高性能计算支持的人工智能应用场景，如大规模文本生成、图像及视频处理等任务。",2,"2026-06-11 03:30:47","trending"]