{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/thunderkittens-simple-fast-and-adorable-ai","title":"ThunderKittens: Simple, Fast, and Adorable AI Kernels","arxiv_id":"2410.20399","date":"2024-10-27","proceeding":null,"authors":["Benjamin F. Spector","Simran Arora","Aaryan Singhal","Daniel Y. Fu","Christopher Ré"],"abstract":"The challenge of mapping AI architectures to GPU hardware is creating a critical bottleneck in AI progress. Despite substantial efforts, hand-written custom kernels fail to meet their theoretical performance thresholds, even on well-established operations like linear attention. The diverse hardware capabilities of GPUs might suggest that we need a wide variety of techniques to achieve high performance. However, our work explores whether a small number of key abstractions can drastically simplify the process. We present ThunderKittens (TK), a framework for writing performant AI kernels while remaining easy to use and maintain. Our abstractions map to the three levels of the GPU hierarchy: (1) at the warp-level, we provide 16x16 matrix tiles as basic data structures and PyTorch-like parallel compute operations over tiles, (2) at the thread-block level, we provide a template for overlapping asynchronous operations across parallel warps, and (3) at the grid-level, we provide support to help hide the block launch and tear-down, and memory costs. We show the value of TK by providing kernels that match or outperform prior kernels for a range of AI operations. We match CuBLAS and FlashAttention-3 on GEMM and attention inference performance and outperform the strongest baselines by $10-40\\%$ on attention backwards, $8\\times$ on state space models, and $14\\times$ on linear attention.","url_abs":"https://arxiv.org/abs/2410.20399v1","url_pdf":"https://arxiv.org/pdf/2410.20399v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"thunderkittens-simple-fast-and-adorable-ai","repo_url":"https://github.com/HazyResearch/ThunderKittens","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":null,"task_name":"GPU"},{"task_slug":"state-space-models","task_name":"State Space Models"}],"methods":[{"method_slug":"attention","method_name":"Attention"},{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2410.20399","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2410.20399"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/HazyResearch/ThunderKittens","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"unverified":21},"by_repo_kind":{"official":{"samples":21,"ran":0,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"5c953e9789e91477","entry":"based_test","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"kernels/based/benchmark.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/kernels/based/benchmark.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"5c953e9789e91477"}},{"code_sha256_prefix":"d0d9b292586f56f3","entry":"compute_twiddle_factors_fft","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"kernels/fftconv/gentests.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/kernels/fftconv/gentests.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"d0d9b292586f56f3"}},{"code_sha256_prefix":"f24f3028b67e35d0","entry":"compute_twiddle_factors_fft","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"kernels/fftconv/gentests_1024.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/kernels/fftconv/gentests_1024.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"f24f3028b67e35d0"}},{"code_sha256_prefix":"54b2749eb0418240","entry":"count_params","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"demos/llama/demo_llama_hf.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/demos/llama/demo_llama_hf.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"54b2749eb0418240"}},{"code_sha256_prefix":"9d855071c55ee252","entry":"fft_matrix","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"kernels/fftconv/gentests.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/kernels/fftconv/gentests.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"9d855071c55ee252"}},{"code_sha256_prefix":"1cf11ad3708c4d6f","entry":"fft_matrix","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"kernels/fftconv/gentests_1024.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/kernels/fftconv/gentests_1024.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"1cf11ad3708c4d6f"}},{"code_sha256_prefix":"7eebf98ddea95c97","entry":"fftconv_tk_test","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"kernels/fftconv/benchmark.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/kernels/fftconv/benchmark.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"7eebf98ddea95c97"}},{"code_sha256_prefix":"2c7a1c21d9730061","entry":"get_based_inputs","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"kernels/based/benchmark.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/kernels/based/benchmark.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"2c7a1c21d9730061"}},{"code_sha256_prefix":"1615a1b5b86e2b15","entry":"get_data","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"demos/based/document_ie_based.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/demos/based/document_ie_based.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"1615a1b5b86e2b15"}},{"code_sha256_prefix":"ed1316b9d8136663","entry":"get_fft_inputs","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"kernels/fftconv/benchmark.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/kernels/fftconv/benchmark.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"ed1316b9d8136663"}},{"code_sha256_prefix":"e12aa72e4c396b45","entry":"get_flops","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"kernels/layernorm/benchmark.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/kernels/layernorm/benchmark.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"e12aa72e4c396b45"}},{"code_sha256_prefix":"7777d7cd21b34e3e","entry":"get_flops","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"kernels/based/benchmark.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/kernels/based/benchmark.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"7777d7cd21b34e3e"}},{"code_sha256_prefix":"8588c8cfb9739504","entry":"get_flops","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"kernels/fftconv/benchmark.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/kernels/fftconv/benchmark.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"8588c8cfb9739504"}},{"code_sha256_prefix":"55de983a3ac2c905","entry":"get_inputs","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"demos/based/document_ie_based.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/demos/based/document_ie_based.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"55de983a3ac2c905"}},{"code_sha256_prefix":"dbe5d402dc4ce748","entry":"get_layer_norm_inputs","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"kernels/layernorm/benchmark.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/kernels/layernorm/benchmark.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"dbe5d402dc4ce748"}},{"code_sha256_prefix":"041f53d114a5e7db","entry":"get_lm_eval_lolcats_model","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"demos/lolcats/demo_lolcats_hf.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/demos/lolcats/demo_lolcats_hf.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"041f53d114a5e7db"}},{"code_sha256_prefix":"7aa5f9a6f612c4e6","entry":"layernorm_test","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"kernels/layernorm/benchmark.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/kernels/layernorm/benchmark.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"7aa5f9a6f612c4e6"}},{"code_sha256_prefix":"457acc5a326ca09e","entry":"pytorch_test","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"kernels/based/gentests.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/kernels/based/gentests.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"457acc5a326ca09e"}},{"code_sha256_prefix":"ddee33c0d89530af","entry":"ref_fftconv","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"kernels/fftconv/gentests.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/kernels/fftconv/gentests.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"ddee33c0d89530af"}},{"code_sha256_prefix":"8733dce8506be321","entry":"ref_fftconv","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"kernels/fftconv/gentests_1024.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/kernels/fftconv/gentests_1024.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"8733dce8506be321"}},{"code_sha256_prefix":"ed9d06db6edca0c4","entry":"setup_fsdp_config","repo":"HazyResearch/ThunderKittens","repo_kind":"official","path":"demos/lolcats/demo_lolcats_hf.py","file_url":"https://github.com/HazyResearch/ThunderKittens/blob/HEAD/demos/lolcats/demo_lolcats_hf.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"ed9d06db6edca0c4"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}