{"name":"basalt","repo":"jesdga95/basalt","url":"https://midorreal.com/project/jesdga95-basalt","repository":"https://github.com/jesdga95/basalt","description":"Local inference engine optimized for NVIDIA Blackwell","facts":{"license":{"value":"MIT","source":"https://github.com/jesdga95/basalt","checked":"22 Nov 2026"},"language":{"value":"C++","source":"https://github.com/jesdga95/basalt","checked":"22 Nov 2026"},"homepage":{"value":"https://basalt.jgulfo.com","source":"https://basalt.jgulfo.com","checked":"22 Nov 2026"},"last_release":{"value":"v0.5.0, 19 Nov 2026","source":"https://github.com/jesdga95/basalt/releases","checked":"22 Nov 2026"},"releases":{"value":"10","source":"https://github.com/jesdga95/basalt/releases","checked":"20 Nov 2026"},"contributors":{"value":"167","source":"https://github.com/jesdga95/basalt/graphs/contributors","checked":"20 Nov 2026"},"archived":{"value":"no","source":"https://github.com/jesdga95/basalt","checked":"22 Nov 2026"}},"stars":1741.0,"writeup":{"what_it_is":"Basalt is a local inference engine for large language models such as Qwen3.8-Flash-Next, DeepSeek-V4.1-Flash and Qwen3.8-27B. It runs on NVIDIA Blackwell GPUs (sm_120) on Linux and Windows x64, and can keep model experts that do not fit in VRAM in system RAM. The server exposes OpenAI and Anthropic compatible APIs.","audience":"Users running large language models locally on NVIDIA Blackwell GPUs","claims":[{"kind":"specific","claim":"Runs on NVIDIA Blackwell GPUs (compute capability 12.0) on Linux and Windows x64","excerpt":"It runs\non an NVIDIA Blackwell GPU (sm_120) on Linux and Windows x64","status":"not_checked"},{"kind":"specific","claim":"Can place experts that do not fit in VRAM into system RAM","excerpt":"with what does not fit in VRAM (the experts of a mixture-of-experts model)\nin system RAM","status":"not_checked"},{"kind":"specific","claim":"Server speaks the OpenAI and Anthropic APIs","excerpt":"The server speaks the OpenAI and Anthropic APIs.","status":"not_checked"},{"kind":"specific","claim":"Supports context up to 1,048,576 tokens for DeepSeek-V4.1-Flash","excerpt":"| Context | up to 262,144 (longer with YaRN) | up to 1,048,576 | up to 262,144 |","status":"not_checked"}],"alternatives":[],"written_by":"AI, from the project's README","date":"2026-11-18"},"why_now":null,"owner_supplied":null,"score":{"verdict":"sleeper","hype":42,"reality":86,"gap":-44,"momentum":null,"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":"2026-11-19T19:10:38+00:00","text":"Latest release v0.5.0","source":"https://github.com/jesdga95/basalt/releases"},{"date":"2026-10-29T15:15:42+00:00","text":"First release","source":"https://github.com/jesdga95/basalt/releases"},{"date":"2026-10-06T22:17:02+00:00","text":"Repository created","source":"https://github.com/jesdga95/basalt"}],"cite":"basalt is a GitHub project at https://github.com/jesdga95/basalt written mainly in C++, described there as \"Local inference engine optimized for NVIDIA Blackwell\".","attribution":"Data from Mid or Real, https://midorreal.com/project/jesdga95-basalt, checked 22 November 2026. Attribution with a link is required.","still_real":null}