Prism ML Website | Whitepaper | Demo & Examples | Colab Notebook | Discord - Deployed footprint — runs on virtually any device - **End-to-end 1-bit weights** across embeddings, attention projections, MLP projections, and LM head - **GGUF Q1_0 (g128)** format for 1-bit packing of weights with shared scales for each group (group size 128). - **Cross-platform**: CUDA (RTX/datacenter), Metal (Mac), Swift (iPhone/iPad), Android - **MLX companion**: also available as MLX 1-bit g128 for native Apple Silicon inference - **Google Colab** — try Bonsai in your browser, no setup required - **Whitepaper** — for more details on Bonsai, check out our whitepaper - **Demo repo** — comprehensive examples for serving, benchmarking, and integrating Bonsai - **Discord** — join the community for support, discussion, and updates - **1-bit kernels**: llama.cpp fork (CUDA + Metal) · MLX fork (Apple Silicon) · mlx-swift fork (iOS/macOS) - **Locally AI** — we have partnered with Locally AI for iPhone support
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Category pending
Input Not verified→ Output Not verified
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Basic profile indexed · More details to follow
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗Prism ML Website | Whitepaper | Demo & Examples | Discord Full 27B-class reasoning in binary transformer weights, for llama.cpp (CUDA, Metal, CPU) > **\~14.2x** smaller than FP16 | **\~90%** of FP16 intelligence retained | **\~44 tok/s** on an Apple M5 Pro laptop
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗Prism ML Website | Whitepaper | Demo & Examples | Discord Full 27B-class reasoning in binary transformer weights — the first 27B-class model to run on a phone > **~14.2x** smaller than FP16 | **~90%** of FP16 intelligence retained | **~11 tok/s** on iPhone 17 Pro Max
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗FP16 safetensors (HuggingFace format) of the 1-bit Bonsai 27B model. This repo exists for users who want to run Bonsai with stock HuggingFace tooling or frameworks that don't yet support 1-bit weights natively. The 1-bit hybrid-attention kernels are currently in our forks of MLX, mlx-swift, and llama.cpp — once they land upstream, this unpacked version will no longer be needed. > **We strongly recommend using the native 1-bit models instead.** The 1-bit format is where all the benefits of Bonsai come from — a 14.2x memory reduction to 3.9 GB, interactive decoding on everyday laptops (44 tok/s on an M5 Pro), and the first 27B-class model that runs on a phone (11 tok/s on iPhone 17 Pro Max). This unpacked FP16 version is full-size (~54 GB) and does not provide any of those advantages.
Category pending
Input Not verified→ Output Not verified
Price not verifiedCheck multimodal and channel pricing separately
View details ↗Prism ML Website | Whitepaper | Demo & Examples | Colab Notebook | Discord - Deployed footprint fits on virtually any device with a GPU - **End-to-end 1-bit weights** across embeddings, attention projections, MLP projections, and LM head - **GGUF Q1_0 (g128)** format for 1-bit packing of weights with shared scales for each group (group size 128). - **Cross-platform**: CUDA (RTX/datacenter), Metal (Mac), Swift (iPhone/iPad), Android - **MLX companion**: also available as MLX 1-bit g128 for native Apple Silicon inference - **Google Colab** — try Bonsai in your browser, no setup required - **Whitepaper** — for more details on Bonsai, check out our whitepaper - **Demo repo** — comprehensive examples for serving, benchmarking, and integrating Bonsai - **Discord** — join the community for support, discussion, and updates - **1-bit kernels**: llama.cpp fork (CUDA + Metal) · MLX fork (Apple Silicon) · mlx-swift fork (iOS/macOS) - **Locally AI** — we have partnered with Locally AI for iPhone support
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Basic profile indexed · More details to follow
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗Prism ML Website | Whitepaper | Demo & Examples | Colab Notebook | Discord > **14.1x** smaller than FP16 | **6.2x** faster on RTX 4090 | **4-5x** lower energy/token - **1.15 GB** parameter memory (down from 16.38 GB FP16) — fits on virtually any device with a GPU - **End-to-end 1-bit weights** across embeddings, attention projections, MLP projections, and LM head - **GGUF Q1_0 (g128)** format with inline dequantization kernels — no FP16 materialization - **Cross-platform**: CUDA (RTX/datacenter), Metal (Mac), Android, CPU - **Competitive benchmarks**: 70.5 avg score across 6 categories, matching full-precision 8B models at 1/14th the size - **MLX companion**: also available as MLX 1-bit g128 for native Apple Silicon inference
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗Prism ML Website | Whitepaper | Demo & Examples | Colab Notebook | Discord > **12.8x** smaller than FP16 | **8.4x** faster on M4 Pro | **44** tok/s on iPhone | runs on Mac, iPhone, iPad Each weight is a single bit: `0` maps to `−scale`, `1` maps to `+scale`. Every group of 128 weights shares one FP16 scale factor.
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗FP16 safetensors (HuggingFace format) of the 1-bit Bonsai-8B model. This repo exists for users who want to run Bonsai with stock HuggingFace tooling or frameworks that don't yet support 1-bit weights natively. The 1-bit kernels are currently in our forks of MLX and llama.cpp — once they land upstream, this unpacked version will no longer be needed. > **We strongly recommend using the native 1-bit models instead.** The 1-bit format is where all the benefits of Bonsai come from — up to 14x memory reduction, 6x faster inference, and 5x lower energy per token. This unpacked FP16 version is full-size and does not provide any of those advantages. - **Bonsai-8B GGUF Q1_0** — 1-bit GGUF for llama.cpp (CUDA, Metal, CPU) - **Bonsai-8B MLX 1-bit** — 1-bit MLX for Apple Silicon
Category pending
Input Not verified→ Output Not verified
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Basic profile indexed · More details to follow
Input Text→ Output Image
Price not verifiedCheck multimodal and channel pricing separately
View details ↗Q2_0 encodes each weight as a 2-bit code `q in {0, 1, 2, 3}`, dequantized via `w = (q - 1) * scale`. One 128-element block is 34 bytes (2 bytes FP16 scale + 32 bytes of packed 2-bit codes) for an effective **2.125 bits/weight**. The fourth code point (`q = 3`, reconstructing to `+2 * scale`) is reserved for future extensions; for ternary weights it is unused. Q2_0 is effectively lossless for ternary weights — the ternary values land exactly on three of the four 2-bit code points.
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。
Category pending
Input Not verified→ Output Not verified
Price not verifiedCheck multimodal and channel pricing separately
View details ↗Bonsai 2 27B is a 27B-parameter reasoning model from PrismML derived from Qwen3.8-27B. It supports coding, mathematics, tool calling, and image understanding with a 262K-token context window. Ternary compression shrinks...
Input Text · Image→ Output Text
$0.075 / M input tokensOpenRouter · prism-ml/ternary-bonsai-2-27b
View details ↗Prism ML Website | Whitepaper | Demo & Examples | Discord Full 27B-class reasoning in ternary transformer weights, for llama.cpp (CUDA, Metal, CPU) > **\~9.3x** smaller than FP16 (ideal) | **98.2%** of FP16 intelligence retained | **\~47 tok/s** on an Apple M5 Max laptop
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗Bonsai 2 needs an activation transform that currently only exists in our llama.cpp fork. Stock llama.cpp will load this file anyway, since it knows the `Q2_0` type and the `qwen35` architecture, and outputs gibberish with no warning. The main repo's `PQ2_0` and `PTQ1_0` use types stock llama.cpp does not know, so it stops with an error rather than running them. That is why they ship there and this one does not. **For normal use:** Ternary-Bonsai-2-27B-gguf, or Ternary-Bonsai-2-27B-mlx-2bit on Apple Silicon. Setup for every backend: Bonsai-demo.
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗Prism ML Website | Whitepaper | Demo & Examples | Discord > **8.60 GB** on disk, language model + vision tower | **98.2%** of FP16 intelligence retained | **\~47 tok/s** on an Apple M5 Max laptop - **8.60 GB** on disk: a 7.67 GB language model (down from \~54 GB FP16) plus the 0.92 GB vision tower. MLX's container stores a scale and a bias per group, so the language model costs 2.25 bits/weight where the same ternary weights take 1.75 in the GGUF PTQ1_0 packing - **98.2% of FP16 intelligence retained**: 84.78 average across 14 thinking-mode benchmarks — far above the conventional IQ2_XXS build (72.59) at less than two-thirds of its footprint, and within 0.4 points of UD-Q4_K_XL at three times the footprint - **Retains thinking, reasoning, and agentic behavior** deep in the sub-4-bit regime, where conventional low-bit representations collapse: math within half a point of full precision (96.57), coding level with the baseline (89.42), agentic tool calling at 74.92 - **End-to-end ternary language weights** across embeddings, attention projections, MLP projections, and LM head, with no high-precision escape hatches behind a low-bit label — a *true* 1.72 bits per weight as a representation, 2.25 as MLX stores it; the vision tower is bundled in this pack, unquantized - **262K-token context** on-device, kept practical by the Qwen3.8-27B hybrid-attention backbone (\~75% linear attention) - **Custom ternary hybrid-attention kernels** on Apple MLX (Python, Swift) and CUDA — packed weights are consumed directly, never expanded back to FP16 - **GGUF companion**: also available as Ternary-Bonsai-2-27B-gguf for llama.cpp (CUDA, Metal, CPU), in two packings — PTQ1_0 (5.95 GB) and PQ2_0 (7.21 GB)
Input Text→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗A 2B-class vision-language model with ternary transformer weights, for llama.cpp (CUDA, Metal, CPU). Bits/weight is file size over the ~1.88B language-model parameters, so it includes the F32 tensors and metadata. The three language-model packs hold the same ternary weights; they differ only in storage layout and kernel speed. > **PrismML-Eng/Bonsai-demo is the source of truth for running these models.** It has the tested setup for every backend and pins a known-good binary release.
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗> **PrismML-Eng/Bonsai-demo is the source of truth for running these models.** Where anything here disagrees with it, it is right. For an OpenAI-compatible server, run `python -m mlx_vlm.server --model bonsai2-2b-mlx` and send the sampling settings in each request. For CUDA, CPU, and llama.cpp on Metal, use the GGUF packs of the same weights: **Ternary-Bonsai-2-2B-gguf**.
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗Prism ML Website | Demo & Examples | Discord A 4B-class vision-language model with ternary transformer weights, for llama.cpp (CUDA, Metal, CPU). - **Ternary language weights** across embeddings, attention projections, MLP projections and the (tied) LM head, stored with a group-wise FP16 scale. - **Three GGUF packings** of the same weights, consumed directly by custom kernels and never expanded back to FP16: **PTQ1_0** (dense trits, 0.95 GB), **PQ2_0** (2-bit slots, 1.15 GB) and **Q2_0** (1.21 GB). - **Vision input** through a separate mmproj pack (Q8_0, or BF16 for reference precision), loaded only when an image arrives. - **262K-token context**, inherited from the Qwen3.5-4B hybrid-attention backbone (24 linear-attention and 8 full-attention layers).
Input Text · Image→ Output Text
Price not verifiedCheck multimodal and channel pricing separately
View details ↗