Reference
Models reference
The models a room can run, where their weights come from, how much memory they need and how much context they get.
Pooled ships no weights. Each device downloads only its share of the layers from Hugging Face by range request, or from another device in the room that has them.
In the room’s picker
Section titled “In the room’s picker”| Model | Quantization | File | Memory the room needs | Context (default / max) |
|---|---|---|---|---|
| Qwen3 1.7B | Q8_0 | 1.83 GB | 4.0 GB | 8192 / 16384 |
| Qwen 3.8 27B | Q4_0 | 16.06 GB | 17.0 GB | 16384 / 32768 |
| Qwen 3.6 35B MoE (35B-A3B) | Q4_0 | 20.84 GB | 22.5 GB | 32768 / 65536 |
“Memory the room needs” is the total across all devices. Raise the context with ?ctx=.
What each one is
Section titled “What each one is”- Qwen3 1.7B. A small dense model, for rooms of phones and light laptops. 28 layers. Its KV cache is f32, about 224 KB per token, so 8k of context is about 1.8 GB. Codex runs on it; Claude Code does not fit, since its prompt alone is about 15k of the 16k.
- Qwen 3.8 27B. A hybrid: 64 layers, 48 Gated DeltaNet and 16 attention. It has a built-in draft layer (MTP) for speculative decoding. Only the attention layers keep a KV cache, 64 KB per token in f16, so 16k of context is about 1 GB.
- Qwen 3.6 35B MoE. The same hybrid layout with 40 layers and 256 experts, 8 active per token (about 3B of the 35B), so it decodes several times faster than the 27B. Its KV cache is 20 KB per token in f16, so 32k is about 0.64 GB. This is the model to use for coding agents.
Weights
Section titled “Weights”| Model | Hugging Face file |
|---|---|
| Qwen3 1.7B | Qwen/Qwen3-1.7B-GGUF Qwen3-1.7B-Q8_0.gguf |
| Qwen 3.8 27B | unsloth/Qwen3.8-27B-GGUF Qwen3.8-27B-Q4_0.gguf |
| Qwen 3.6 35B MoE | bartowski/Qwen_Qwen3.6-35B-A3B-GGUF Qwen_Qwen3.6-35B-A3B-Q4_0.gguf |
The Qwen3 dense models also fetch config.json and tokenizer.json from the matching Qwen/Qwen3-* repository. The 27B and the MoE read everything from the GGUF.
Answer length
Section titled “Answer length”| Setting | Tokens |
|---|---|
| Short | 150 |
| Normal | 400 |
| Long | 1200 |
| With thinking on | 1200, shared by the think block and the answer |
A prompt must leave at least 32 tokens for the answer. Generation stops before the KV cache would overflow. The Serve API has its own limits: 16384 output tokens by default, capped by the room’s context.
KV cache format
Section titled “KV cache format”f16 by default. With ?kv=q8 on the host, the 27B and the MoE keep it as int8 (about 56% of the memory); the host sends the choice to every device.
Models for tests and development
Section titled “Models for tests and development”These stay in room/models.js for tests. The picker shows them with ?dev=1.
| Model | Format | Engine | Memory | Context |
|---|---|---|---|---|
| Qwen3 0.6B | GGUF Q8_0 | DenseEngine |
0.8 GB | 2048 |
| Qwen3 4B | GGUF Q8_0 | DenseEngine |
4.6 GB | 2048 |
| SmolLM2 135M | safetensors bf16 | DenseEngine |
0.6 GB | 2048 |
Qwen3 0.6B is the golden-test model. Qwen3.5-122B-A10B MoE runs in the engine and its tests but is not in the room.
Engines
Section titled “Engines”| Engine | Models |
|---|---|
Qwen35Engine |
Qwen 3.8 27B, Qwen 3.6 35B MoE (qwen35moe), Qwen3.5-122B-A10B |
DenseEngine |
Qwen3 0.6B, 1.7B, 4B, SmolLM2 135M |
Both are Pooled’s own WebGPU engines. See Engine and kernels.
Adding a model
Section titled “Adding a model”Models with the hybrid layout of Qwen 3.5, 3.6 and 3.8 need only entries in room/models.js. See Adding a model.