llms.txt
Compare with other sites' llms.txt files →# Mirai Labs
> On-device AI stack for Apple Silicon: the uzu inference engine, Mirai-quantized open models, and a local chat app for macOS. Benchmarks against MLX and llama.cpp are published for every uzu release.
Site: https://trymirai.com. Contact: contact@trymirai.com.
## Benchmarks
- [uzu vs MLX vs llama.cpp: Apple Silicon LLM Benchmarks 2026](https://trymirai.com/metrics): uzu 0.6.15 vs MLX 0.32.4 vs llama.cpp on 11 Macs, 24 quantized models: output and input tok/s and memory. Benchmarked 22 Oct 2026. Engines: uzu 0.6.15, MLX 0.32.4, llama.cpp 0.8.0-dev+8642541b2, MTPLX 2.12.2, oMLX 0.7.2, Splash 1.5.0, MLX Swift 3.32.3, llama.cpp 0.7.0, tensorfold 1.1.9.
## Models
Full per-model data (SDK identifiers, install commands, per-device benchmarks): https://trymirai.com/llms-full.txt
### Qwen3.5
- [Qwen3.5 0.8B Mirai-M 4-bit](https://trymirai.com/models/alibaba-qwen3-5-0-8b-mirai-mirai-m-4): Mirai-M 4-bit, 800M parameters, 460 MB, 262144 context
- [Qwen3.5 0.8B Mirai-L 8-bit](https://trymirai.com/models/alibaba-qwen3-5-0-8b-mirai-mirai-l-8): Mirai-L 8-bit, 800M parameters, 824 MB, 262144 context
- [Qwen3.5 2B Mirai-M 4-bit](https://trymirai.com/models/alibaba-qwen3-5-2b-mirai-mirai-m-4): Mirai-M 4-bit, 2B parameters, 1.1 GB, 262144 context
- [Qwen3.5 2B Mirai-L 8-bit](https://trymirai.com/models/alibaba-qwen3-5-2b-mirai-mirai-l-8): Mirai-L 8-bit, 2B parameters, 2.0 GB, 262144 context
- [Qwen3.5 4B Mirai-M 4-bit](https://trymirai.com/models/alibaba-qwen3-5-4b-mirai-mirai-m-4): Mirai-M 4-bit, 4B parameters, 2.9 GB, 262144 context
- [Qwen3.5 4B Mirai-L 8-bit](https://trymirai.com/models/alibaba-qwen3-5-4b-mirai-mirai-l-8): Mirai-L 8-bit, 4B parameters, 4.9 GB, 262144 context
- [Qwen3.5 9B Mirai-M 4-bit](https://trymirai.com/models/alibaba-qwen3-5-9b-mirai-mirai-m-4): Mirai-M 4-bit, 9B parameters, 6.0 GB, 262144 context
- [Qwen3.5 9B Mirai-L 8-bit](https://trymirai.com/models/alibaba-qwen3-5-9b-mirai-mirai-l-8): Mirai-L 8-bit, 9B parameters, 10.0 GB, 262144 context
### Qwen3.6
- [Qwen3.6 27B Mirai-M 4-bit](https://trymirai.com/models/alibaba-qwen3-6-27b-mirai-mirai-m-4): Mirai-M 4-bit, 27B parameters, 15.6 GB, 262144 context
- [Qwen3.6 27B Mirai-L 8-bit](https://trymirai.com/models/alibaba-qwen3-6-27b-mirai-mirai-l-8): Mirai-L 8-bit, 27B parameters, 28.8 GB, 262144 context
### Qwen3.8
- [Qwen3.8 27B Mirai-M 4-bit](https://trymirai.com/models/alibaba-qwen3-8-27b-mirai-mirai-m-4): Mirai-M 4-bit, 27B parameters, 15.7 GB, 262144 context
- [Qwen3.8 27B Mirai-L 8-bit](https://trymirai.com/models/alibaba-qwen3-8-27b-mirai-mirai-l-8): Mirai-L 8-bit, 27B parameters, 28.9 GB, 262144 context
### LFM2.5
- [LFM2.5 230M Mirai-M 4-bit](https://trymirai.com/models/liquidai-lfm2-5-230m-mirai-mirai-m-4): Mirai-M 4-bit, 230M parameters, 130 MB, 128000 context
- [LFM2.5 230M Mirai-L 8-bit](https://trymirai.com/models/liquidai-lfm2-5-230m-mirai-mirai-l-8): Mirai-L 8-bit, 230M parameters, 243 MB, 128000 context
- [LFM2.5 350M Mirai-M 4-bit](https://trymirai.com/models/liquidai-lfm2-5-350m-mirai-mirai-m-4): Mirai-M 4-bit, 350M parameters, 198 MB, 128000 context
- [LFM2.5 350M Mirai-L 8-bit](https://trymirai.com/models/liquidai-lfm2-5-350m-mirai-mirai-l-8): Mirai-L 8-bit, 350M parameters, 373 MB, 128000 context
- [LFM2.5 1.2B Instruct Mirai-M 4-bit](https://trymirai.com/models/liquidai-lfm2-5-1-2b-instruct-mirai-mirai-m-4): Mirai-M 4-bit, 1.2B parameters, 639 MB, 128000 context
- [LFM2.5 1.2B Thinking Mirai-M 4-bit](https://trymirai.com/models/liquidai-lfm2-5-1-2b-thinking-mirai-mirai-m-4): Mirai-M 4-bit, 1.2B parameters, 639 MB, 128000 context
- [LFM2.5 1.2B Instruct Mirai-L 8-bit](https://trymirai.com/models/liquidai-lfm2-5-1-2b-instruct-mirai-mirai-l-8): Mirai-L 8-bit, 1.2B parameters, 1.2 GB, 128000 context
- [LFM2.5 1.2B Thinking Mirai-L 8-bit](https://trymirai.com/models/liquidai-lfm2-5-1-2b-thinking-mirai-mirai-l-8): Mirai-L 8-bit, 1.2B parameters, 1.2 GB, 128000 context
- [LFM2.5 2.6B Mirai-M 4-bit](https://trymirai.com/models/liquidai-lfm2-5-2-6b-mirai-mirai-m-4): Mirai-M 4-bit, 2.6B parameters, 1.5 GB, 131072 context
- [LFM2.5 2.6B Mirai-L 8-bit](https://trymirai.com/models/liquidai-lfm2-5-2-6b-mirai-mirai-l-8): Mirai-L 8-bit, 2.6B parameters, 2.8 GB, 131072 context
### Muse-Glimmer
- [Muse-Glimmer 30B Mirai-M 4-bit](https://trymirai.com/models/meta-muse-glimmer-30b-mirai-mirai-m-4): Mirai-M 4-bit, 30B parameters, 16.6 GB, 131072 context
- [Muse-Glimmer 30B Mirai-L 8-bit](https://trymirai.com/models/meta-muse-glimmer-30b-mirai-mirai-l-8): Mirai-L 8-bit, 30B parameters, 30.3 GB, 131072 context
### d3-flash
- [d3-flash 9B Mirai-M 4-bit](https://trymirai.com/models/vllm-sr-d3-flash-9b-mirai-mirai-m-4): Mirai-M 4-bit, 9B parameters, 4.3 GB
## Product
- [Local LLM library for Apple Silicon](https://trymirai.com/models): 25 Qwen, LFM, Muse-Glimmer and d3-flash checkpoints in 4-bit (Mirai-M) and 8-bit (Mirai-L) for Apple Silicon, with measured uzu tok/s.
- [uzu: On-device LLM inference engine for iPhone, iPad & Mac](https://trymirai.com/inference-runtime): uzu runs LLMs on iPhone, iPad and Mac faster than MLX and llama.cpp. Benchmarks published for each release.
- [Conversion and optimization toolkit](https://trymirai.com/conversion-optimization-toolkit): Convert and optimize your model for iPhone, iPad and Mac. One command to get your model running on 2 billion Apple devices.
- [Mirai for Mac: Local AI Chat on Apple Silicon](https://trymirai.com/chat-for-mac): Chat with your favorite AI models directly on your Mac. Privately and securely. Built natively for macOS and Apple Silicon.
- [Developers](https://trymirai.com/developers): Run a recommended model on your Mac, compare uzu with other engines, and integrate it with the SDK, the Mac app or a coding agent.
## Research
- [The Computer Should Know the Way](https://trymirai.com/blog/the-computer-should-know-the-way): A research essay from Mirai Labs: why the layer that turns what I want into actions should run on my own Mac, call operations directly, and answer while the thought is still there.
- [Tackling random foreign characters](https://trymirai.com/blog/tackling-random-foreign-characters): We randomly saw some Chinese and Russian characters in the middle of coherent English outputs, while benchmarking our new dflashtfm release. The benchmark runs 20 MT-Bench prompts through uzu at temperature 1 with no top-p or top-k, and generates tens of thousands of tokens in total.
- [Speculative decoding in uzu](https://trymirai.com/blog/speculative-decoding-in-uzu): We are releasing our speculative decoding implementation in Uzu. It supports Qwen3.6 27B, Qwen3.8 27B and Muse Glimmer. On Apple M5-series chips, we outperform MTPLX by almost 2x, and llama.cpp by over 3x at comparable quantization levels, with the strongest gains acheived on mathematical reasoning and coding tasks.
- [Trees from Marginals: Autoregressive drafting with factorized priors](https://arxiv.org/abs/2607.06763): Weaver, a lightweight autoregressive adapter that constructs proposal trees from the top-K marginals of a factorized drafter.
- [Sparse Buffers for KV Cache](https://trymirai.com/blog/sparse-buffers-for-kv-cache): How uzu backs the KV cache with Metal sparse buffers so memory grows page by page instead of committing 4.5 GiB up front for a 32k-context Qwen3-4B.
- [Introducing Mirai Quantization: Redefining the speed-quality frontier for local LLMs on Apple silicon.](https://trymirai.com/blog/quantization): Mirai-M (4-bit) and Mirai-L (8-bit): a co-designed quantization and inference stack with 40-60% more tokens per second than llama.cpp and MLX at the same quality level.
## Company
- [On-Device AI for Apple Silicon: uzu & Models | Mirai Labs](https://trymirai.com/): Mirai Labs builds uzu, a Rust LLM engine for iPhone, iPad and Mac, and 4-bit/8-bit Qwen, LFM, Muse-Glimmer and d3-flash models benchmarked vs MLX and llama.cpp.
- [About Mirai Labs: the On-Device AI Lab for Apple Silicon](https://trymirai.com/about-us): We are the frontier on-device AI lab, building the full on-device stack to achieve realtime local intelligence.
- [Careers](https://trymirai.com/careers): Join a small, senior team building the full on-device stack to achieve realtime local intelligence.
- [Contact Mirai Labs: uzu, models, Mac app & SDK](https://trymirai.com/i-am-interested): Contact Mirai Labs. Email contact@trymirai.com or use the form to tell us what interests you in Mirai.
- [Member of Technical Staff](https://trymirai.com/careers/member-of-technical-staff-model-optimization-and-post-training): Apply for the member of technical staff role at Mirai. Join a small, senior team building the full on-device stack to achieve realtime local intelligence.
- [Inference engineer](https://trymirai.com/careers/inference-engineer): Apply for the inference engineer role at Mirai. Join a small, senior team building the full on-device stack to achieve realtime local intelligence.
## Links
- [Docs](https://docs.trymirai.com/)
- [uzu on GitHub](https://github.com/trymirai/uzu)
- [lalamo on GitHub](https://github.com/trymirai/lalamo)
- [X (Twitter)](https://x.com/trymirai)
- [GitHub](https://github.com/trymirai)
- [Hugging Face](https://huggingface.co/trymirai)
- [LinkedIn](https://www.linkedin.com/company/trymirai)
- [Discord](https://discord.com/invite/trymirai)