# AI use cases for Janas (prabanta-dev/janas) Repo: [prabanta-dev/janas](https://github.com/prabanta-dev/janas) | Use cases: 5 | Page: https://dedupedai.com/use-cases/prabanta-dev-janas/ Janas is a family of C tools whose first component, Janas-LLM, runs large mixture-of-experts language models on ordinary computers, with or without a GPU. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Chat with an 80B model on a 32 GB laptop: https://dedupedai.com/use-cases/prabanta-dev-janas/run-large-model-on-laptop-locally/ - Measure your own machine's speed with janas-bench: https://dedupedai.com/use-cases/prabanta-dev-janas/benchmark-own-machine-and-send-report/ - Chat with a local model on a Windows laptop: https://dedupedai.com/use-cases/prabanta-dev-janas/chat-with-local-model-on-windows-laptop/ - Cap memory and keep the GPU off on a shared machine: https://dedupedai.com/use-cases/prabanta-dev-janas/limit-memory-and-disable-gpu-on-shared-machine/ - Speed up chat replies with draft tokens: https://dedupedai.com/use-cases/prabanta-dev-janas/speed-up-chat-replies-with-drafts/ ## How to chat with an 80B model on a 32 GB laptop with Janas Repo: [prabanta-dev/janas](https://github.com/prabanta-dev/janas) | Difficulty: Intermediate | Time: 1 hour | Basis: From the README Install Janas-LLM on Linux and chat with a model larger than the laptop's RAM, offline after download. **Who it is for:** A developer or researcher with a 32 GB Linux laptop who wants a capable local chat model without a cloud account. **The problem:** Most local model runners need the whole model in memory, so a model bigger than the laptop's RAM cannot run at all. **The outcome:** A working chat session with a locally run model, where the weights stay on disk and expert weights are streamed from the SSD as needed. ### Steps 1. Download the Linux build: `curl -LO https://github.com/prabanta-dev/janas/releases/latest/download/janas-linux-x86_64.tar.gz` 2. Unpack it and enter the folder: `tar xf janas-linux-x86_64.tar.gz && cd janas` 3. Start the chat: `./janas-chat` 4. Pick a model from the list shown; it is downloaded and checked before use 5. Type a question and read the reply; press Ctrl-C to stop a reply and Ctrl-D to leave the chat ### Prompt to paste into your agent ```text In the janas repo at , run ./janas-chat, choose the Qwen3-Next-80B-A3B model, and ask it one short question. Report the reply text and the tokens-per-second figure if shown. Verify that the process finished without the desktop becoming unresponsive. ``` ### Caveat The README says the engine is deliberately hard on the machine and is not verified; on a machine with less headroom the desktop can be pushed into swap and become slow or unresponsive. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/prabanta-dev-janas/run-large-model-on-laptop-locally/ ## How to measure your own machine's speed with janas-bench with Janas Repo: [prabanta-dev/janas](https://github.com/prabanta-dev/janas) | Difficulty: Intermediate | Time: 1 hour | Basis: From the README Run janas-bench on your hardware to get tokens-per-second figures and a report for the project. **Who it is for:** A hardware-curious user or contributor with a CPU or GPU the README did not test, such as a different laptop or an AMD machine. **The problem:** Published speeds come from one laptop, so it is unclear how a different CPU, GPU, or disk will perform. **The outcome:** Your own reading and writing speeds for a model, plus a benchmark report you can share with the project. ### Steps 1. Complete the install from the Try it section and make sure `./janas-chat` runs once 2. Run the benchmark: `janas-bench --prompt 1024 --gen 16` 3. For a steadier figure, run three rounds: `janas-bench --rounds 3` 4. Compare the reading and writing tokens-per-second with the README table 5. Send the report from janas-bench back as described in Helping it run on more hardware ### Prompt to paste into your agent ```text In the janas repo at , run janas-bench --prompt 1024 --gen 16 and then janas-bench --rounds 3. Output the reading and writing tokens-per-second figures for each run. Verify that both runs completed and that the figures are reported as numbers, not errors. ``` ### Caveat The README says figures on another machine will differ and that the benchmark may need several passes before the engine settles on its choices, so a first run is usually slower. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/prabanta-dev-janas/benchmark-own-machine-and-send-report/ ## How to chat with a local model on a Windows laptop with Janas Repo: [prabanta-dev/janas](https://github.com/prabanta-dev/janas) | Difficulty: Beginner | Time: 45 minutes | Basis: From the README Unpack the Windows build and chat with a local model from a terminal. **Who it is for:** A Windows 11 user who wants a private local chat model and is not a developer. **The problem:** Running a large local model normally needs tooling and setup that a typical Windows user does not have. **The outcome:** A chat session with a locally run model on a Windows x86-64 machine. ### Steps 1. Download janas-windows-x86_64.zip from the Windows section of the README page 2. Unpack the zip file to a folder you can find 3. Open a terminal in the Janas folder inside the unpacked files 4. Start the chat: `.\janas-chat` 5. Pick a model from the list; it is downloaded and checked before use ### Prompt to paste into your agent ```text On a Windows x86-64 machine, unpack janas-windows-x86_64.zip into , open a terminal in the janas folder, and run .\janas-chat. Report which models are offered and confirm that a chat prompt appears. ``` ### Caveat The README measured the Windows build only on one laptop under Windows 11, so other Windows machines are untested and may be slower or refuse a model. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/prabanta-dev-janas/chat-with-local-model-on-windows-laptop/ ## How to cap memory and keep the GPU off on a shared machine with Janas Repo: [prabanta-dev/janas](https://github.com/prabanta-dev/janas) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Start Janas-LLM with a memory limit or without the GPU so it does not take over a shared computer. **Who it is for:** A user on a shared or work machine who wants a local model without it filling the memory or touching the GPU. **The problem:** The engine fills free memory with its expert cache and can use the GPU, which may disturb other work on the same machine. **The outcome:** A chat session that keeps to a set memory limit and runs on the CPU only. ### Steps 1. Run the chat with the GPU off (inferred): `JANAS_GPU=0 ./janas-chat` (inferred, check the README) 2. To cap memory, start the chat with the `--cache` option set to a limit of your choice (inferred flag form, check the value the README gives in docs/chat.md) 3. Watch the desktop responsiveness during the first replies 4. Stop a reply with Ctrl-C if the machine slows down ### Prompt to paste into your agent ```text In the janas repo, run JANAS_GPU=0 ./janas-chat and ask one short question. Report the reply and confirm from the output that no GPU was used. Verify that the process ended cleanly after Ctrl-D. ``` ### Caveat The README warns that using a GPU has once caused a driver reset that took the desktop with it, so keeping the GPU off is the safer choice on a machine you cannot afford to lose. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/prabanta-dev-janas/limit-memory-and-disable-gpu-on-shared-machine/ ## How to speed up chat replies with draft tokens with Janas Repo: [prabanta-dev/janas](https://github.com/prabanta-dev/janas) | Difficulty: Intermediate | Time: 1 hour | Basis: From the README Enable draft decoding for supported Qwen and Gemma models to get faster chat replies. **Who it is for:** A user chatting with a Qwen3.5, Qwen3.6, Qwen3-Next, or Gemma 4 model who wants faster answers to ordinary questions. **The problem:** Writing a reply one token at a time is slow on a model this large, even when the machine can run it. **The outcome:** Chat replies at a higher rate, with the same text, when the model supports drafts. ### Steps 1. Install and start the chat as in the Try it section 2. Choose a supported model such as Qwen3-Next-80B-A3B or Gemma-4-26B-A4B-it 3. Turn drafts on for the chat (inferred flag form): `./janas-chat --mtp` (inferred, check the README) 4. Ask an ordinary English question and time the reply 5. Compare the tokens-per-second with and without drafts ### Prompt to paste into your agent ```text In the janas repo, start ./janas-chat --mtp with Qwen3.6-35B-A3B selected and ask it to explain how bread is made at home. Report the tokens-per-second shown and confirm the reply is complete. ``` ### Caveat The README says the draft gain depends on the text, and on Italian prose one model kept only about a third of its drafts and gained little. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/prabanta-dev-janas/speed-up-chat-replies-with-drafts/