GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
webbrain-one
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
webbrain-one
Model Tree
moonshotai/Kimi-K2.6
Input Modalities
Output Modalities
Supported Functionality
At WebBrain, we build browser agents that need to understand the visual state of the web—not just extracted text. Screenshots, charts, dashboards, rich editors, and the location and appearance of controls are part of real browser work, so vision is a practical product requirement.
In our American–Chinese open-model frontier benchmark, Laguna XS 2.1 stood out as a strong model in its size class, but the upstream checkpoint is text-only. This project adds a basic MoonViT vision bridge while keeping both the language backbone and vision tower frozen.
[!IMPORTANT] The pinned BF16 text backbone, frozen MoonViT tower, and final 100K-example projector are complete, packaged, and fingerprint-verified. Multimodal model code, processor/serving integration, and final image-inference validation are still pending, so this is not yet a stock ready-to-serve checkpoint.
| Component | Detail |
|---|---|
| Text backbone | Laguna XS 2.1, 33B total / 3B active MoE, 2048 hidden size — frozen |
| Packaged text weights | BF16 from poolside/Laguna-XS-2.1 — exact pinned copy |
| Vision tower | MoonViT-3d from Kimi-K2.6, 27 layers, 1152-dimensional patch features — frozen |
| Projector | LayerNorm -> 2x2 merge -> Linear(4608, 4608) -> GELU -> Linear(4608, 2048) — trained in BF16 |
| Projector size | 30,679,808 trainable parameters |
| Training envelope | Up to 512 merged image tokens inside 2,048-token training sequences |
| Backbone context | 262,144 tokens, inherited from Laguna XS 2.1 |
| Target hardware | Blackwell; final single-RTX PRO 6000 96 GB serving recipe pending validation |
| File | Contents | Size | SHA-256 |
|---|---|---|---|
vision_tower.safetensors | Frozen MoonViT-3d tower, 329 tensors, all BF16 | 833,765,768 bytes | befe801bd7dfe8bf5630fef56a7f53c2235065599ca9eea4d995040e2e6fd183 |
mm_projector.safetensors | Final step-782 PatchMerger projector, 6 tensors, all BF16 | 61,360,104 bytes | 7837384f18be69a4f875ca44a8ed69ec186501d70896daf4b305fa77547974de |
Machine-readable provenance is in
VISION_ADAPTER_MANIFEST.json.
The packaged text backbone is copied from
poolside/Laguna-XS-2.1 at immutable revision
e9df9a59996d790b94b70f3fef343fe1d9e34bdf.
Every copied file was checked against its upstream Git blob or LFS SHA-256
fingerprint, then independently rechecked after upload.
Projector training used the frozen BF16 Laguna XS 2.1 backbone and frozen MoonViT-3d features. This repository now pairs the final trained projector and frozen tower with the verified BF16 backbone above. End-to-end behavior will be validated after the remaining multimodal integration is assembled.
A serving command is intentionally not published yet. The current repository contains the complete text backbone, frozen vision tower, and final projector, but not the multimodal processor and serving assembly that connects them. A tested quickstart will be added after integration passes the final gates.
The overall construction and model-card approach was inspired by Baseten's GLM-5.2-Vision-NVFP4: keep the text backbone and MoonViT tower frozen, train a compact PatchMerger projector between them, and publish the provenance and hardware constraints explicitly. Credit to the Baseten team for demonstrating this practical recipe. This repository does not reuse Baseten model weights, benchmark results, or deployment artifacts.
The redistributed Laguna XS 2.1 BF16 backbone remains subject to the included OpenMDW-1.1 license and Poolside's source notices. The included MoonViT tower remains subject to the included Kimi-K2.6 Modified MIT terms. The newly trained projector is documented in the manifest above. Downstream users remain responsible for complying with both upstream licenses.
Built on Poolside's Laguna XS 2.1 BF16 and Moonshot AI's Kimi-K2.6, with the vision-attachment method inspired by Baseten's GLM-5.2-Vision-NVFP4. These teams were not involved in this development checkpoint; please do not direct issues with this repository to them.
Ask your inference provider—such as OpenRouter or another OpenAI-compatible managed service—to deploy this exact repository with its multimodal processor and serving plugin. Deploying only the upstream text backbone will not enable image input.
[!WARNING] Experimental vision package. The MoonViT adapter and its serving package are experimental. Live end-to-end validation of the NVFP4 package on an RTX 5090 is still provisioning: the pinned SGLang container image is being fetched, so neither the text nor image smoke test has passed yet. Fine-grained OCR, small-object and control identification, GUI grounding, and hallucination calibration remain limited. The roadmap is to expand and diversify the training data, add more high-resolution OCR and UI examples, and pursue broader parameter-efficient tuning if community interest warrants the investment. To contribute evaluation or training data, sponsor compute, or explore a design partnership, use the community interest form. Do not use this package as the sole decision source for safety-critical automation.