{"name":"janas","repo":"prabanta-dev/janas","url":"https://midorreal.com/project/prabanta-dev-janas","repository":"https://github.com/prabanta-dev/janas","description":"37 tok/s from Qwen3-Next-80B (48 GB) on a 32 GB laptop, 24 without drafts vs 2.8 for llama.cpp; reading 62 vs 16 tok/s. Faster than llama.cpp on all 13 models compared in Q4_K_M. LLMs bigger than your RAM, pure C, CPU or GPU.","facts":{"license":{"value":"GPL-3.0","source":"https://github.com/prabanta-dev/janas","checked":"22 Nov 2026"},"language":{"value":"C","source":"https://github.com/prabanta-dev/janas","checked":"22 Nov 2026"},"last_release":{"value":"v2026.11.19.3, 19 Nov 2026","source":"https://github.com/prabanta-dev/janas/releases","checked":"22 Nov 2026"},"releases":{"value":"32","source":"https://github.com/prabanta-dev/janas/releases","checked":"21 Nov 2026"},"contributors":{"value":"1","source":"https://github.com/prabanta-dev/janas/graphs/contributors","checked":"21 Nov 2026"},"archived":{"value":"no","source":"https://github.com/prabanta-dev/janas","checked":"22 Nov 2026"}},"stars":742.0,"writeup":{"what_it_is":"Janas is a family of C tools for running language models on a local computer, using its CPU, Vulkan GPU, Intel NPU and NVMe SSD. It can run models larger than RAM by streaming expert weights from disk, and includes a terminal chat, a desktop app, an HTTP server, a downloader and a benchmark tool.","audience":"Developers and users who want to run large language models on their own consumer hardware.","claims":[{"kind":"specific","claim":"Qwen3-Next-80B-A3B Q4_K_M writes at 26.4 tok/s on the development laptop, against 3.0 tok/s for llama.cpp under the same conditions","excerpt":"**Qwen3-Next-80B-A3B** Q4_K_M, 48 GB on 32 GB of memory: writing a reply","status":"not_checked"},{"kind":"specific","claim":"Runs models larger than RAM by reading the experts each token needs from the NVMe SSD","excerpt":"when it does not - a mixture of experts larger than the RAM - the experts each token needs are read from the NVMe SSD","status":"not_checked"},{"kind":"specific","claim":"Written in C with no dependencies beyond the C library","excerpt":"written in C with no dependencies beyond the C library","status":"not_checked"},{"kind":"specific","claim":"Runs on Linux x86-64 and Windows x86-64","excerpt":"Linux x86-64 (Ubuntu 22.04, Debian 12 and later)","status":"not_checked"},{"kind":"specific","claim":"Provides an HTTP API compatible with OpenAI's, Ollama's and Anthropic's interfaces","excerpt":"an HTTP API that follows OpenAI's, Ollama's and Anthropic's (Claude Code)","status":"not_checked"}],"alternatives":[],"written_by":"AI, from the project's README","date":"2026-11-17"},"why_now":null,"owner_supplied":null,"score":{"verdict":"real","hype":60,"reality":62,"gap":-2,"momentum":0.4,"confidence":60,"version":"score-2.0","rule":"verdict-2.0","calculated_at":"2026-11-22T08:00:00+00:00","verdict_source":"rule"},"early_read":null,"changes":[{"date":"2026-11-19T18:11:07+00:00","text":"Latest release v2026.11.19.3","source":"https://github.com/prabanta-dev/janas/releases"},{"date":"2026-10-11T00:30:21+00:00","text":"First release","source":"https://github.com/prabanta-dev/janas/releases"},{"date":"2026-09-20T09:55:09+00:00","text":"Repository created","source":"https://github.com/prabanta-dev/janas"}],"cite":"janas is a GitHub project at https://github.com/prabanta-dev/janas written mainly in C, described there as \"37 tok/s from Qwen3-Next-80B (48 GB) on a 32 GB laptop, 24 without drafts vs 2.8 for llama.cpp; reading 62 vs 16 tok/s. Faster than llama.cpp on all 13 models compared in Q4_K_M. LLMs bigger than your RAM, pure C, CPU or GPU\".","attribution":"Data from Mid or Real, https://midorreal.com/project/prabanta-dev-janas, checked 22 November 2026. Attribution with a link is required.","still_real":null}