GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
ukisai
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
ukisai
Model Tree
ornith-ai/Ornith-1.5-9B
Input Modalities
Output Modalities
Supported Functionality
We built Swift by identifying reasoning-marker tokens that, in our analysis, trigger overthinking in the model's reasoning rollouts. We then fine-tuned the model by penalizing usage of those tokens while it reasons.
Swift produces shorter reasoning traces while keeping accuracy close to the base model. In our testing, Swift 1.5 also hits the output cap less often than the base on every benchmark below.
The vision encoder, the multi-token prediction (MTP) layer, the token embeddings and the output head are unchanged from Ornith-1.5-9B.
All scores compare Ornith-1.5-9B with Swift 1.5 Ornith 9B, both in BF16 with thinking on. Both models used the same harness, sampling, seeds and output cap on every row. Token columns report thinking tokens per attempt; the visible answer is not counted.
Swift 1.5 is ahead of base on GPQA-Diamond (+1.52 pp). It trails base on AIME 2024 (−0.67 pp), AIME 2026 (−2.00 pp), IFBench (−2.13 pp) and LiveCodeBench v6 (−2.29 pp).
This is an experimental research release. Swift 1.5 Ornith 9B thinks less than the base on every benchmark above, but it also scores 0.7 to 2.3 pp lower than the base on AIME, IFBench and LiveCodeBench. Only GPQA-Diamond, the benchmark we used to select this release, is higher.
We have not evaluated it on agentic, tool-use or vision benchmarks, and benchmark scores may not translate into reliable general-purpose behavior. These observations concern the specific model, runtimes and tests we used. This release shares a research result and its current limitations, rather than presenting a production-ready model.
The BF16 checkpoint is about 19 GB and runs on a single GPU. For the full 262,144-token context, use an 80 GB GPU. It needs the same runtimes as Ornith-1.5-9B: Transformers ≥ 5.8.1, vLLM ≥ 0.19.1 or SGLang ≥ 0.5.9.
The model thinks before it answers. With the qwen3 reasoning parser below, the server returns the thinking in
reasoning_content and the answer in content. We evaluated with temperature=1.0, top_p=0.95, top_k=20, min_p=0.0.
GGUF files for llama.cpp are in ukisai/Swift-1.5-Ornith-9B-GGUF.
bash
vllm serve ukisai/Swift-1.5-Ornith-9B \ --dtype bfloat16 \ --max-model-len 262144 \ --enable-prefix-caching \ --reasoning-parser qwen3 \ --enable-auto-tool-choice \ --tool-call-parser qwen3_xml \ --port 8000bash
python -m sglang.launch_server \ --model-path ukisai/Swift-1.5-Ornith-9B \ --dtype bfloat16 \ --context-length 262144 \ --reasoning-parser qwen3 \ --tool-call-parser qwen3_coder \ --port 8000Both commands start an OpenAI-compatible server on port 8000. Lower the context length if your GPU has less memory. For contexts beyond 262,144 tokens (YaRN), see the Ornith-1.5-9B model card.
python
import torchfrom transformers import AutoModelForImageTextToText, AutoProcessor model_id = "ukisai/Swift-1.5-Ornith-9B"processor = AutoProcessor.from_pretrained(model_id)model = AutoModelForImageTextToText.from_pretrained( model_id, torch_dtype=torch.bfloat16, device_map="auto",)The checkpoint includes the base model's one-layer MTP head. To enable self-speculative decoding, append the corresponding flags to the server command above:
bash
# vLLM--speculative-config '{"method":"mtp","num_speculative_tokens":3}' # SGLang--speculative-algorithm NEXTN --speculative-num-steps 3 \ --speculative-eagle-topk 1 --speculative-num-draft-tokens 4Swift 1.5 Ornith 9B is a derivative of Ornith-1.5-9B (Ornith Team, MIT License). UkisAI's contribution, including the adapted weights, is licensed under the Apache License 2.0. See NOTICE for the change notice and attribution details.
The base model's MIT License still applies to the base model. This is an independent UkisAI release, not an official Ornith release or endorsement.
bibtex
@misc{swift-1.5-ornith-9b, title = {Swift 1.5 Ornith 9B}, author = {UkisAI}, year = {2026}, url = {https://huggingface.co/ukisai/Swift-1.5-Ornith-9B}}We acknowledge Amazon Web Services for providing compute credits and infrastructure support for Swift's development, training, and evaluation.
vllm serve ukisai/Swift-1.5-Ornith-9B \ --dtype bfloat16 \ --max-model-len 262144 \ --enable-prefix-caching \ --reasoning-parser qwen3 \ --enable-auto-tool-choice \ --tool-call-parser qwen3_xml \ --port 8000python -m sglang.launch_server \ --model-path ukisai/Swift-1.5-Ornith-9B \ --dtype bfloat16 \ --context-length 262144 \ --reasoning-parser qwen3 \ --tool-call-parser qwen3_coder \ --port 8000import torchfrom transformers import AutoModelForImageTextToText, AutoProcessor model_id = "ukisai/Swift-1.5-Ornith-9B"processor = AutoProcessor.from_pretrained(model_id)model = AutoModelForImageTextToText.from_pretrained( model_id, torch_dtype=torch.bfloat16, device_map="auto",)# vLLM--speculative-config '{"method":"mtp","num_speculative_tokens":3}' # SGLang--speculative-algorithm NEXTN --speculative-num-steps 3 \ --speculative-eagle-topk 1 --speculative-num-draft-tokens 4@misc{swift-1.5-ornith-9b, title = {Swift 1.5 Ornith 9B}, author = {UkisAI}, year = {2026}, url = {https://huggingface.co/ukisai/Swift-1.5-Ornith-9B}}