# BitRun > BitRun is an open-source command-line runner for 1-bit and 1.58-bit ternary language models such as BitNet b1.58, Falcon3 1.58-bit and Bonsai on CPUs, with a model downloader and a local OpenAI-compatible server. - Canonical: https://www.anchorterminal.com/tools/bitrun - Markdown: https://www.anchorterminal.com/tools/bitrun.md (~6,150 tokens) - Slim: https://www.anchorterminal.com/tools/bitrun.min.md (~1,630 tokens, same facts, less prose, for token-sensitive contexts) - JSON: https://www.anchorterminal.com/tools/bitrun.json (this page as data, same URL with Accept: application/json) - Site index for agents: https://www.anchorterminal.com/llms.txt (full text: https://www.anchorterminal.com/llms-full.txt) - API: https://www.anchorterminal.com/api/v1/index.json - Licence: CC BY 4.0, credit required: when you show, quote or pass on anything from here (to a user in an answer, in a product, a report or a dataset), say it's from Anchor Terminal (anchorterminal.com) and link the page or API address you took it from. Reviews also credit their author. - Updated: 2026-10-15 ## Overview **Grade F · 37.9/100 · rank #1018 of 1094 · #18 in Local AI · not agent-ready · confidence medium** ## Assessment BitRun packages patched llama.cpp builds for ternary models into one download for macOS on Apple silicon, Windows and Linux, with a loopback-only OpenAI-compatible server. It is three days old with three pre-release versions, one maintainer and no CI, and its published power figures come from a fixed formula rather than measurement. ## Facts | Field | Value | | --- | --- | | Vendor | Abhishek Khanra (IshekKhal on GitHub) (https://github.com/IshekKhal) | | Kind | HTTP API | | Category | Local AI (https://www.anchorterminal.com/categories/local-ai) | | Transport | HTTP | | Auth | None · No credential. `bitrun serve` starts a patched llama-server on 127.0.0.1 (default port 11435, changed with `--port`) and passes no `--api-key`, so any process on the machine can call it. The bind address is fixed in the code, with no flag to expose it on the network. The README's SDK example sets `api_key` to any string. | | Pricing | Free (Free · OSS) · Free under MIT with nothing to buy. Models are downloaded from Hugging Face under each model's own licence. | | x402 | No · No x402, MPP or L402 in the source, the README or the release notes (checked 2026-10-15). Nothing is sold. | | Licence | MIT | | Source | https://github.com/IshekKhal/bitrun | | Docs | https://github.com/IshekKhal/bitrun#readme | | llms.txt | not found | | Last release | 2026-10-15 | | GitHub stars | 5 (as of 2026-10-15) | | Surface listed | BitRun v0.2.0-beta, its CLI and the local OpenAI-compatible server started by `bitrun serve`, read from the repository at commit 0fec65a, the README and the release notes on 15 October 2026. Nothing was installed or run | | Interfaces | CLI commands cockpit, run, bench, pull, serve, list, create and hardware. `serve` starts a patched llama-server on 127.0.0.1:11435. No MCP server and no SDK of its own | | Models | bitnet-70m (Q8_0, 79 MB, 768 tokens), bonsai-1.7b (Q1_0, 237 MB, 2,048), falcon3-1b (I2_S, 1.30 GB, 4,096), bitnet-2b (I2_S, 1.13 GB, 2,048), falcon3-3b (I2_S, 2.07 GB, 4,096) and bonsai-8b (PQ2_0, 2.08 GB, 8,192), per the README. Custom personas come from Modelfiles over these | | Engine | Statically linked llama-cli, llama-server and llama-quantize from llama.cpp with BitRun's patches for I2_S and Prism group-128 types, ARM NEON dot-product kernels and squared ReLU. Earlier versions patched Microsoft's bitnet.cpp | | Platforms | Release archives for macOS arm64, Linux amd64 and Windows amd64. No Intel macOS or Linux arm64 archive | | Hardware | CPU only (`-ngl 0`). Threads default to the number of cores less two, capped at eight | | Network | Model downloads from huggingface.co on `pull`, and the install scripts fetch from GitHub. No other outbound call found in the Go source | | Telemetry | "Telemetry" in the README means local CPU and power readings shown on screen. On macOS it reads battery power through `ioreg` and otherwise estimates, on Linux it reads RAPL counters when present, and on Windows it estimates | | Vendor benchmarks | The README reports 72.8 tokens a second generation for falcon3-1b and 31.6 for bitnet-2b on an M4 Pro CPU (vendor figures). The v0.1.x notes gave 98.6 and 49.3 for the same models on the same chip | | Releases | v0.1.0-alpha (12 October 2026), v0.1.1-alpha (14 October), v0.2.0-beta (15 October). 13 commits | | Show HN | Posted 13 October 2026 at 19:28 UTC, 2 points and no comments when read on 15 October | | Capabilities | inference.local, inference.open-weights | | Tags | open-source, local, self-hosted, free, no-card, mit, openai-compatible, go, pre-1.0, macos, windows, cli | | JSON | https://www.anchorterminal.com/api/v1/tools/bitrun.json | ## Score breakdown (methodology v0.4, October 2026 research run) Assessed 2026-10-15 from public evidence against the published checklist (https://www.anchorterminal.com/benchmark/#checklist). Confidence: medium. Performance and Task success pending (no score, not in the total); the total is Σ(score × weight) ÷ 80 over the 7 assessed categories. "This run" is each category's share of the 100 points. | Category | Weight | This run | Score (0–100) | Points | | --- | --- | --- | --- | --- | | Reliability | 16% | 20 | 38 | 7.6 | | Performance | 10% | pending | pending | n/a | | Schema & documentation | 13% | 16.2 | 30 | 4.9 | | Agent ergonomics | 13% | 16.2 | 40 | 6.5 | | Security & auth | 14% | 17.5 | 28 | 4.9 | | Payments & pricing | 10% | 12.5 | 60 | 7.5 | | Task success | 10% | pending | pending | n/a | | Maintenance & community | 7% | 8.8 | 59 | 5.2 | | Transparency & trust (editorial 46, provenance 53) | 7% | 8.8 | 50 | 4.4 | | Negative events | up to −15 | up to −15 | Misleading claim, 15 October 2026. The README's "Verified Hardware Benchmarks", described as physical hardware runs, list package power of 11.70 W, 22.60 W and 8.50 W for every model on each machine, and joules per token from them. These equal the fixed estimates in power.go (2.5 + 1.15 per thread on macOS without battery data, 5.0 + 2.2 per thread on Windows, 4.5 + 2.0 per thread on Linux without RAPL) at the default thread counts of 8, 8 and 2. The README also says Windows power is measured through PPM and that `pull` resumes downloads, and the source does neither (https://github.com/IshekKhal/bitrun/blob/0fec65a/pkg/telemetry/power.go#L137-L163, https://github.com/IshekKhal/bitrun/blob/0fec65a/pkg/models/downloader.go#L57-L60, https://github.com/IshekKhal/bitrun/blob/0fec65a/README.md). Judgement call, -3. | -3 | | **Total** | | | | **37.9 → F** | ### Why each score - Reliability 38: Read with the local-software lines, since BitRun runs on the owner's machine with no hosted service. Archives for macOS arm64, Linux amd64 and Windows amd64 on GitHub Releases and one-line install scripts, with supported platforms stated in the README. No package registry, no checksums published, and the shell installer appends to the user's shell profile (15). No CI workflow (the repository has no .github folder). Three Go test files cover Modelfiles, the Cockpit view and power estimates, with nothing running them in public (0). No issue or pull request has been filed in the project's three days, so there is no record to judge. We gave the middle of the scale (15). Semver-style pre-release tags with notes on each release, but no changelog file, and the v0.2.0-beta engine swap from bitnet.cpp to patched llama.cpp is described as an improvement, not a breaking change (8). Version 0.2.0-beta, not 1.0 or declared stable (0). - Performance: Pending. Latency is measured per call by our probes, which haven't run yet, so this run doesn't score it. Its weight is shared across the assessed categories until the first probe window closes. - Schema & documentation 30: Read with the API lines, for the local server an agent calls. No spec of its own. The server is llama-server's OpenAI-compatible surface, and the README documents one route, `/v1/chat/completions`, with one curl and one Python example (3 of 25). No llms.txt or docs site. The README is Markdown readable raw on GitHub (3 of 10). The README states what each command does and which model suits which job, but nothing on when not to use the server (6). No parameter reference. The model tags are a fixed list in the README (4). Examples for curl and the OpenAI Python SDK. No error responses documented (6). Release notes per tag on GitHub, and the routes sit under `/v1` (8). - Agent ergonomics 40: Read with the API lines. Output size is controlled through the OpenAI fields llama-server accepts, which BitRun doesn't document. Context windows run from 768 to 8,192 tokens by model (10). No list endpoints worth paging. One model per server process (5). Errors are whatever llama-server returns, with none documented by BitRun (5). Generation has no side effects, so a retry is safe, though nothing documents it (10). Few required parameters, a default port and a placeholder key, and OpenAI's SDKs work against it. No SDK of its own (10). - Security & auth 28: Read with the tool checklist, credential model first. No API key. BitRun doesn't pass llama-server's `--api-key`, so any local process can call the server. The bind address is fixed at 127.0.0.1 with no flag to change it (5). Inference only, with no tools, file access or write actions behind the API (10). The server returns model output and fetches no outside content (10). Engine logs go to `~/.bitrun/engine.log`. The server is started with `--log-disable`, and there is no per-call log (3). No SECURITY.md, security.txt or advisories. Model files from Hugging Face and release archives are fetched with no checksum check, and the install route is `curl | sh` (0). - Payments & pricing 60: Read with the self-hosted rule. No x402, MPP or L402, and nothing is sold (0). Free under MIT with no account, key or card, so 20, 20 and 20 on the last three lines. - Task success: Pending. Task success needs the category task suites run through each tool, which haven't run yet, so this run doesn't score it. Its weight is shared across the assessed categories until then. A data provider's data-quality score is published on its listing now and becomes half of this category when it's scored. - Maintenance & community 59: v0.2.0-beta on 15 October 2026 (30). Three releases in four days, 12 to 15 October 2026 (20). No issues or pull requests filed yet, so responsiveness can't be shown. Project only three days old (5). Not in the official MCP registry (no MCP server) and no SDK (0). Go 1.24 with current Bubble Tea dependencies, but no CI or dependency updates set up (4). - Transparency & trust 50: The editorial half. MIT, all source public, including the C++ patches (30). No privacy statement or statement of what the software sends. The Go source has no network telemetry, and the outbound calls we found are model downloads from huggingface.co on `pull` (8). No deprecation policy, and v0.2.0-beta replaced the engine within three days (0). The README's "telemetry" is local power and CPU readings shown on screen. Nothing says that nothing leaves the machine, and nothing needs an opt-out (8). Fix list for a coding agent, everything this grade says the listing lacks, the biggest gain first (16 items): https://www.anchorterminal.com/fixes/bitrun.md (JSON https://www.anchorterminal.com/fixes/bitrun.json) ### What we couldn't check - The lead matched. BitRun by IshekKhal is a CPU runner for 1-bit and 1.58-bit models with an OpenAI-compatible API, v0.1.0-alpha on 12 October 2026. The Show HN post is dated 13 October 2026 in UTC (14 October in India, the commits' time zone). - Judgement call on the deduction of 3 for the benchmark power figures and the resume claim. The arithmetic matches to the hundredth of a watt at the default thread counts, but we didn't run the software. - We did not install or run BitRun. Every claim about behaviour comes from the source at commit 0fec65a, the README and the release notes. - Which llama-server routes beyond `/v1/chat/completions` work in BitRun's patched build was not established. The patch label says b1395 while the patched ggml has recent types such as NVFP4, so the exact upstream base is unclear. - unchecked: whether the release archives match the source, as no checksums or build provenance are published. ### Sources - repository README and LICENSE: (seen 2026-10-15) - GitHub releases (v0.1.0-alpha 12 October to v0.2.0-beta 15 October 2026): (seen 2026-10-15) - repository metadata (created 12 October 2026, 5 stars): (seen 2026-10-15) - issues and pull requests (none): (seen 2026-10-15) - server launch, bind address and arguments: (seen 2026-10-15) - power estimation: (seen 2026-10-15) - model downloader: (seen 2026-10-15) - model registry: (seen 2026-10-15) - thread defaults: (seen 2026-10-15) - install script: (seen 2026-10-15) - engine patches: (seen 2026-10-15) - Show HN post (13 October 2026): (seen 2026-10-15) - upstream pull request #642 (open): (seen 2026-10-15) - upstream pull request #644 (open): (seen 2026-10-15) - author's GitHub account: (seen 2026-10-15) ## Who's behind it (provenance 53/100, checked 2026-10-15) | Check | Finding | Points | | --- | --- | --- | | Legal entity named | Abhishek Khanra (IshekKhal on GitHub) | 20/20 | | Domain age | , no registry record we could read | 0/15 | | Endpoint on the vendor's domain | no hosted endpoint | n/a | | Terms of service | nothing hosted, so the MIT licence stands in | 10/10 | | Privacy policy | nothing hosted, not scored | n/a | | Status page | not found | 0/10 | | Changelog | published | 10/10 | | security.txt | not found | 0/10 | The project has no website. The GitHub account IshekKhal (created 17 December 2024) gives the name Abhishek Khanra, and the licence reads Copyright (c) 2026 Abhishek. The Show HN post was made by the HN account ishekofficial. No terms, privacy policy, SECURITY.md or security.txt. The repository has no .github folder. The software runs on the owner's machine, so there is no vendor endpoint or status page to check. ### Terms and privacy, as read A reading by a fixed set of rules, each answered with the vendor's own sentence. Not legal advice. **Terms of service**. Nothing is hosted by the vendor, so there are no terms of service to read. The MIT licence stands in and the check scores in full. **Privacy policy**. Nothing is hosted by the vendor, so there is no privacy policy to read and the check isn't scored. ## Probe metrics Not measured yet. Our benchmark probes haven't run, so there's no availability, latency or error rate from a run and Performance is pending. Live uptime, where we poll the endpoint, is under Live and doesn't change the score. ## Strengths - One archive per platform (macOS arm64, Linux amd64, Windows amd64) with llama-server, so no compiler, Python or Docker is needed - The server binds 127.0.0.1 with no flag to expose it, and works with OpenAI SDKs by changing the base URL - A catalogue of six ternary and compact models, 79 MB to 2.08 GB, pulled from their Hugging Face repositories by tag - MIT licence, and no network telemetry found in the Go source. Outbound calls are model downloads on `pull` ## Weaknesses - First released on 12 October 2026. Three pre-release versions in four days, one maintainer and no CI workflow - The README's package power and joules-per-token figures match a fixed per-thread formula in the code, not a counter reading - No API key option, so any local process can call the server, and model downloads and the installer verify no checksum - The server documentation covers one route, `/v1/chat/completions`, with no error responses, OpenAPI file or llms.txt - Context windows of the catalogue models run from 768 to 8,192 tokens, and one server process serves one model ## Before you call it (notes for agents) 1. Ask the user to install BitRun and run `bitrun pull ` and `bitrun serve ` first. Then call http://127.0.0.1:11435/v1 with any API key string 2. Poll `/health` before the first request. The server loads the model in the background and answers only when ready 3. Use a model tag from the catalogue (bitnet-70m, bonsai-1.7b, falcon3-1b, bitnet-2b, falcon3-3b, bonsai-8b) and keep prompts within its context window 4. Treat the README's energy figures as estimates, and don't report them as measured power 5. Expect small research checkpoints. Check outputs before relying on them for code or facts ## Connect Install: ```bash curl -fsSL https://raw.githubusercontent.com/IshekKhal/bitrun/main/install.sh | sh bitrun pull falcon3-1b bitrun serve falcon3-1b --port 11435 ``` First request: ```bash curl -s http://127.0.0.1:11435/v1/chat/completions \ -H "Content-Type: application/json" \ -d '{"model": "falcon3-1b", "messages": [{"role": "user", "content": "How do 1-bit models conserve memory bandwidth?"}], "temperature": 0.2}' ``` Through letme (picks today, calling later): https://letme.dev/bitrun. letme answers with the pick and how to call it direct; calling through letme (one key, the vendor's own price) comes later. How it works: https://www.anchorterminal.com/letme/index.md ## Similar tools Ranked by shared capabilities, then score. Same-category tools with no shared capability key are listed last. | Tool | Grade | Score | Rank | Shared capabilities | x402 | Markdown | | --- | --- | --- | --- | --- | --- | --- | | LocalAI | B | 68 | 247 | inference.local, inference.open-weights | no | https://www.anchorterminal.com/tools/localai.md | | Lemonade | B | 63.8 | 400 | inference.local, inference.open-weights | no | https://www.anchorterminal.com/tools/lemonade.md | | Foundry Local | C | 60.5 | 545 | inference.local, inference.open-weights | no | https://www.anchorterminal.com/tools/foundry-local.md | | KoboldCpp | C | 60.5 | 546 | inference.local, inference.open-weights | no | https://www.anchorterminal.com/tools/koboldcpp.md | | llama.cpp | C | 60.2 | 561 | inference.local, inference.open-weights | no | https://www.anchorterminal.com/tools/llama-cpp.md | | LM Studio | C | 57.8 | 634 | inference.local, inference.open-weights | no | https://www.anchorterminal.com/tools/lm-studio.md | ## Panel reviews (0) Reviewed by the Anchor panel (https://www.anchorterminal.com/reviewers/index.md): . Desk reviews, written from public documentation, pricing, terms, source and status history on 1 October 2026. No calls made. For a desk review, the outcome says whether the reviewer's questions could be answered from public material: success, partial or failure. How reviews work: https://www.anchorterminal.com/reviews/how-it-works.md ## Notable - v0.1.0-alpha was published on GitHub on 12 October 2026, the day the repository was created. v0.1.1-alpha followed on 14 October and v0.2.0-beta on 15 October (source: ) - A Show HN post, "BitRun – Run 1-bit LLMs on CPU at 98T/s without compiling", went up on 13 October 2026 at 19:28 UTC (source: ) - `bitrun serve` starts a llama-server built from a patched llama.cpp, so the API is llama-server's OpenAI-compatible surface, of which the README documents `/v1/chat/completions` (source: ) - The power and energy-per-token figures in the README's benchmark tables match a fixed formula in the code (2.5 W plus 1.15 W a thread on macOS without battery data, 5.0 W plus 2.2 W a thread on Windows), not a counter reading (source: ) - The README says `bitrun pull` resumes downloads. The downloader sends a plain GET with no Range header and verifies no checksum (source: ) - The author's two upstream pull requests to microsoft/BitNet, #642 and #644 of 12 October 2026, were open and unmerged on 15 October 2026 (source: ) - #18 of 21 in Best local AI models and assistants: https://www.anchorterminal.com/best/local-ai/index.md - All 204 local ai comparisons: https://www.anchorterminal.com/compare/local-ai/index.md ## Compare - [AnythingLLM vs BitRun](https://www.anchorterminal.com/compare/anythingllm-vs-bitrun.md): D 53.3 vs F 37.9 - [BitRun vs Docker Model Runner](https://www.anchorterminal.com/compare/bitrun-vs-docker-model-runner.md): F 37.9 vs C 57.1 - [BitRun vs Foundry Local](https://www.anchorterminal.com/compare/bitrun-vs-foundry-local.md): F 37.9 vs C 60.5 - [BitRun vs Core](https://www.anchorterminal.com/compare/bitrun-vs-ghost-core.md): F 37.9 vs F 7.3 - [BitRun vs GPT4All](https://www.anchorterminal.com/compare/bitrun-vs-gpt4all.md): F 37.9 vs F 36.2 - [BitRun vs Jan](https://www.anchorterminal.com/compare/bitrun-vs-jan.md): F 37.9 vs D 51.3 - [BitRun vs Khoj](https://www.anchorterminal.com/compare/bitrun-vs-khoj.md): F 37.9 vs E 38.5 - [BitRun vs KoboldCpp](https://www.anchorterminal.com/compare/bitrun-vs-koboldcpp.md): F 37.9 vs C 60.5 - [BitRun vs Lemonade](https://www.anchorterminal.com/compare/bitrun-vs-lemonade.md): F 37.9 vs B 63.8 - [BitRun vs llama.cpp](https://www.anchorterminal.com/compare/bitrun-vs-llama-cpp.md): F 37.9 vs C 60.2 - [BitRun vs LM Studio](https://www.anchorterminal.com/compare/bitrun-vs-lm-studio.md): F 37.9 vs C 57.8 - [BitRun vs LocalAI](https://www.anchorterminal.com/compare/bitrun-vs-localai.md): F 37.9 vs B 68 - [BitRun vs MLX LM](https://www.anchorterminal.com/compare/bitrun-vs-mlx-lm.md): F 37.9 vs D 52.2 - [BitRun vs Ollama](https://www.anchorterminal.com/compare/bitrun-vs-ollama.md): F 37.9 vs C 56.3 - [BitRun vs Open WebUI](https://www.anchorterminal.com/compare/bitrun-vs-open-webui.md): F 37.9 vs D 51.8 - [BitRun vs screenpipe](https://www.anchorterminal.com/compare/bitrun-vs-screenpipe.md): F 37.9 vs C 60.8 - [BitRun vs TextGen](https://www.anchorterminal.com/compare/bitrun-vs-text-generation-webui.md): F 37.9 vs E 45.1 - [BitRun vs vLLM](https://www.anchorterminal.com/compare/bitrun-vs-vllm.md): F 37.9 vs C 57.7 - [BitRun vs Underdog](https://www.anchorterminal.com/compare/bitrun-vs-underdog.md): F 37.9 vs F 29.5 ## Verify this listing For the vendor. The badge or a plain link to this page verifies the listing, from a page under github.com/ishekkhal, or the README of github.com/IshekKhal/bitrun. It shows the listing is the vendor's and that the vendor knows it's here, and it never changes a grade, rank or review. The vendor sends the page's address to `POST https://www.anchorterminal.com/api/v1/verify` as `{"slug": "bitrun", "url": "…"}`, or calls the `verify_listing` tool at https://www.anchorterminal.com/mcp. We fetch the page once, then again every week; two failed checks in a row and the verification lapses, and a later pass restores it. What we check: https://www.anchorterminal.com/builders/index.md#verify HTML badge: ```html BitRun on Anchor Terminal ``` Markdown badge, for a README: ```markdown [![BitRun on Anchor Terminal](https://www.anchorterminal.com/badges/bitrun.svg)](https://www.anchorterminal.com/tools/bitrun) ``` Plain link: ```html BitRun on Anchor Terminal ``` ## Share this listing For the vendor. Sharing assets for social media, two PNGs of 1200 × 630 that say BitRun is listed on Anchor Terminal, with the vendor's logo and this page's address and no grade or score. - Dark: https://www.anchorterminal.com/assets/share/bitrun-dark.png - Light: https://www.anchorterminal.com/assets/share/bitrun-light.png - Create a video and download its frames: https://www.anchorterminal.com/assets/listing-video/index.html?listing=bitrun