https://pypi.org/project/turbo-attn/ Recent updates to the Python Package Index for turbo-attn en https://pypi.org/project/turbo-attn/0.6.4/ Optimized CUDAgraph-enabled kernels and attention backend for vLLM, SGLang and more based on TurboQuant near-lossless KV cache compression. SOTA performance with Gemma 4, Qwen 3.6 and other modern LLMs. dmitri.evseev@arbi.city Sat, 13 Jun 2026 14:11:48 GMT https://pypi.org/project/turbo-attn/0.6.3/ Optimized CUDAgraph-enabled kernels and attention backend for vLLM, SGLang and more based on TurboQuant near-lossless KV cache compression. SOTA performance with Gemma 4, Qwen 3.6 and other modern LLMs. dmitri.evseev@arbi.city Sat, 13 Jun 2026 00:11:03 GMT https://pypi.org/project/turbo-attn/0.6.2/ Optimized CUDAgraph-enabled kernels and attention backend for vLLM, SGLang and more based on TurboQuant near-lossless KV cache compression. SOTA performance with Gemma 4, Qwen 3.6 and other modern LLMs. dmitri.evseev@arbi.city Fri, 12 Jun 2026 21:54:02 GMT https://pypi.org/project/turbo-attn/0.6.1/ Optimized CUDAgraph-enabled kernels and attention backend for vLLM, SGLang and more based on TurboQuant near-lossless KV cache compression. SOTA performance with Gemma 4, Qwen 3.6 and other modern LLMs. dmitri.evseev@arbi.city Fri, 12 Jun 2026 19:43:12 GMT https://pypi.org/project/turbo-attn/0.6.0/ Optimized CUDAgraph-enabled kernels and attention backend for vLLM, SGLang and more based on TurboQuant near-lossless KV cache compression. SOTA performance with Gemma 4, Qwen 3.6 and other modern LLMs. dmitri.evseev@arbi.city Fri, 12 Jun 2026 15:19:31 GMT https://pypi.org/project/turbo-attn/0.5.1/ Optimized CUDAgraph-enabled kernels and attention backend for vLLM, SGLang and more based on TurboQuant near-lossless KV cache compression. SOTA performance with Gemma 4, Qwen 3.6 and other modern LLMs. dmitri.evseev@arbi.city Fri, 12 Jun 2026 15:15:25 GMT https://pypi.org/project/turbo-attn/0.5.0/ Optimized CUDAgraph-enabled kernels and attention backend for vLLM, SGLang and more based on TurboQuant near-lossless KV cache compression. SOTA performance with Gemma 4, Qwen 3.6 and other modern LLMs. dmitri.evseev@arbi.city Thu, 11 Jun 2026 15:03:08 GMT https://pypi.org/project/turbo-attn/0.4.1/ Optimized CUDAgraph-enabled kernels and attention backend for vLLM, SGLang and more based on TurboQuant near-lossless KV cache compression. SOTA performance with Gemma 4, Qwen 3.6 and other modern LLMs. dmitri.evseev@arbi.city Thu, 11 Jun 2026 05:09:05 GMT https://pypi.org/project/turbo-attn/0.4.0/ Optimized CUDAgraph-enabled kernels and attention backend for vLLM, SGLang and more based on TurboQuant near-lossless KV cache compression. SOTA performance with Gemma 4, Qwen 3.6 and other modern LLMs. dmitri.evseev@arbi.city Wed, 10 Jun 2026 21:20:08 GMT https://pypi.org/project/turbo-attn/0.3.2/ Optimized CUDAgraph-enabled kernels and attention backend for vLLM, SGLang and more based on TurboQuant near-lossless KV cache compression. SOTA performance with Gemma 4, Qwen 3.6 and other modern LLMs. dmitri.evseev@arbi.city Wed, 10 Jun 2026 20:43:15 GMT https://pypi.org/project/turbo-attn/0.3.0/ Optimized CUDAgraph-enabled kernels and attention backend for vLLM, SGLang and more based on TurboQuant near-lossless KV cache compression. SOTA performance with Gemma 4, Qwen 3.6 and other modern LLMs. dmitri.evseev@arbi.city Wed, 10 Jun 2026 19:09:17 GMT https://pypi.org/project/turbo-attn/0.3.1/ Optimized CUDAgraph-enabled kernels and attention backend for vLLM, SGLang and more based on TurboQuant near-lossless KV cache compression. SOTA performance with Gemma 4, Qwen 3.6 and other modern LLMs. dmitri.evseev@arbi.city Wed, 10 Jun 2026 19:09:14 GMT https://pypi.org/project/turbo-attn/0.2.0/ Optimized CUDAgraph-enabled kernels and attention backend for vLLM, SGLang and more based on TurboQuant near-lossless KV cache compression. SOTA performance with Gemma 4, Qwen 3.6 and other modern LLMs. dmitri.evseev@arbi.city Tue, 02 Jun 2026 19:33:14 GMT https://pypi.org/project/turbo-attn/0.1.2/ Optimized CUDAgraph-enabled kernels and attention backend for vLLM, SGLang and more based on TurboQuant near-lossless KV cache compression. SOTA performance with Gemma 4, Qwen 3.6 and other modern LLMs. dmitri.evseev@arbi.city Thu, 30 Apr 2026 20:20:38 GMT https://pypi.org/project/turbo-attn/0.1.1/ Optimized CUDAgraph-enabled kernels and attention backend for vLLM, SGLang and more based on TurboQuant near-lossless KV cache compression. SOTA performance with Gemma 4, Qwen 3.6 and other modern LLMs. dmitri.evseev@arbi.city Thu, 30 Apr 2026 19:49:21 GMT https://pypi.org/project/turbo-attn/0.1.0/ Productionized TurboQuant KV cache compression for vLLM and SGLang (imported as `tqkv`) dmitri.evseev@arbi.city Thu, 30 Apr 2026 19:41:29 GMT