JiRackUltra_1b
A fast and efficient ~1.5B model optimized for CPU inference. The model was refactored with BitNet features and an updated tokenizer that includes new Routing, Tool call, and Robotics tags. Built on a redesigned DeepSeek R1 architecture with native ternary (BitNet-style) support and ready-to-run GGUF quantizations. - JiRack is a cloud-ready model that helps save money on cloud infrastructure. It can be used as an exp...
Params
1.8 B
Context
131,072
Downloads 30d
2.5 M
Likes
14
Download history
daily snapshots · 29 days2.5 M307 K
Sep 25Oct 4Oct 14Oct 23
Can you run it?
Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.
| File | Quant | Size | Est. VRAM | Verdict on RTX 4090 · 24 GB |
|---|---|---|---|---|
| JiRackUltra_1b_Q3_K_M.gguf | Q3_K_M | 0.9 GB | 1.8 GB | ✅ Runs comfortably |
| JiRackUltra_1b_Q4_K_M.gguf | Q4_K_M | 1.1 GB | 2.0 GB | ✅ Runs comfortably |
| JiRackUltra_1b.Q6_K.gguf | Q6_K | 1.5 GB | 2.4 GB | ✅ Runs comfortably |
| JiRackUltra_1b.Q8_0.gguf | Q8_0 | 1.9 GB | 2.9 GB | ✅ Runs comfortably |
| JiRackUltra_1b.gguf | GGUF | 3.6 GB | 4.7 GB | ✅ Runs comfortably |
| model.safetensors | bf16 | 3.6 GB | 4.7 GB | ✅ Runs comfortably |
| model.safetensors (bf16, full) | bf16 + 128K ctx | 3.6 GB | 8.7 GB | ✅ Runs comfortably |
Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.
Run it
copy-paste, exact tags checked against the Hub$ ollama run jirackultra:1b # pin the quantization explicitly $ ollama run jirackultra:1b-q4_k_m
$ huggingface-cli download CMSManhattan/JiRackUltra_1b-GGUF \
JiRackUltra_1b_Q4_K_M.gguf --local-dir .
$ llama-cli -m JiRackUltra_1b_Q4_K_M.gguf \
-c 8192 -ngl 99 -t 8 --color
$ curl -s https://aimodelscomparison.com/api/v1/models/jirackultra-1b
{
"hf_id": "CMSManhattan/JiRackUltra_1b",
"params_b": 1.78,
"context_length": 131072,
"license": { "id": "mit", "commercial": "yes" },
"downloads_30d": 2485543,
"vram_estimates": [
{ "quant": "Q6_K", "gb": 2.4 },
{ "quant": "Q8_0", "gb": 2.9 }
],
"updated_at": "2026-10-16T01:00:04Z"
}
Specifications
- Architecture
- Qwen2ForCausalLM
- Parameters
- 1.8 B
- Tensor type
- BF16
- Context length
- 131,072
- Vocabulary
- 151,936
- Layers / heads
- 28 / 12
- Licence
- mit
- First seen on the Hub
- 2026-08-02
- Training datasets
- undisclosed
- Added to our catalog
- 2026-09-25
Compare with any text-generation model
Popular comparisons
Appears in