---
license: other
license_name: swift-open-license-1.0
license_link: https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b/blob/main/LICENSE
library_name: gguf
pipeline_tag: image-text-to-text
tags:
- gguf
- llama.cpp
- qwen3_8
- reasoning
- efficient-thinking
- token-efficient
- post-training
- terminal-bench
base_model: ukisai/Swift-1.5-Qwen3.8-27b
base_model_relation: quantized
---
# Swift 1.5 Qwen3.8-27B
**GGUF quantizations.** Derived directly from [Swift 1.5](https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b) with [llama.cpp](https://github.com/ggml-org/llama.cpp). Run a chosen quantization tier with a current llama.cpp-compatible runtime such as `llama-server`.
Swift 1.5 Qwen3.8-27B is UkisAI's reasoning-efficient derivative of [Qwen3.8-27B](https://huggingface.co/Qwen/Qwen3.8-27B).
It uses **58.5% fewer thinking tokens** while scoring **0.35% higher** than the base, for a **9.18× speed-up** on several tasks.
Swift 1.5 is a direct upgrade from [Swift 1.0](https://huggingface.co/ukisai/Swift-Qwen3.8-27b), our model with 350k+ downloads, delivering stronger overall performance than both base and Swift 1.0 in various tasks, especially coding and agentic, while using fewer thinking tokens. We accomplished that by scaling up the post-training (RL and OPD) from the previous version.
## Demo
We gave base Qwen3.8-27B and Swift 1.5 27B the same prompt:
> create a 3d little planet globe where I (player can walk around) and it has all these biomes to explore, the globe doesn't have to be too big, but still fun to go around. It's about a boy scout who is camping and goes around exploring.
Try the game yourself here: [https://ukisai.com/swift-games/27b](https://ukisai.com/swift-games/27b)
Base Qwen3.8-27B took 104.6 minutes to build its game. Swift 1.5 took 11.39 minutes.
## Training approach
We made Swift efficient by figuring out which tokens were linked to pathological overthinking and penalizing them without "attacking" the reasoning length directly then regained the accuracy with RL and OPD, leading to "compressed" token usage while maintaining accuracy.
Swift 1.5 was made from [Swift 1.0](https://huggingface.co/ukisai/Swift-Qwen3.8-27b), on whom we scaled up the post-training methods that previously improved Swift1.0 model performance, this time with the main
focus on long-horizon, agentic, and coding tasks, as seen in the LiveCodeBench and Terminal Bench 2.1 improvements. Our training data is viewable here: https://huggingface.co/datasets/ukisai/Qwen3.8-27B-multi-turn-agent-sft albeit it is not used out of the box, but rather re-sampled, turned into proper RL environments etc.
## Evaluation
The external results below compare **Qwen3.8-27B**, the foundation base model,
and **Swift 1.5**. Both models use
the same saved evaluation protocols, and all scores are reported as final aggregate
percentages.
| Benchmark |
Final score |
Mean tokens |
Median tokens |
| Qwen3.8 |
Swift 1.5 |
Qwen3.8 |
Swift 1.5 |
Reduction |
Reduction |
| General reasoning |
| GPQA-Diamond | 88.28% | 88.59% | 15,014 | 8,717 | ↓ 41.9% | ↓ 58.5% |
| C-Eval | 90.00% | 90.92% | 1,492 | 819 | ↓ 45.1% | ↓ 16.9% |
| IFBench | 73.53% | 72.07% | 8,052 | 4,955 | ↓ 38.5% | ↓ 47.3% |
| ERQA | 67.45% | 65.40% | 4,137 | 1,906 | ↓ 53.9% | ↓ 56.2% |
| Mathematics |
| AIME 2026 | 98.67% | 96.00% | 22,014 | 13,203 | ↓ 40.0% | ↓ 48.5% |
| HMMT November 2025 | 99.33% | 97.33% | 22,032 | 14,957 | ↓ 32.1% | ↓ 47.8% |
| Coding |
| LiveCodeBench v6 | 76.76% | 81.71% | 11,184 | 8,448 | ↓ 24.5% | ↓ 46.3% |
| Agent tasks |
| Terminal-Bench 2.1 | 69.21% | 72.13% | 52,265 | 43,733 | ↓ 16.3% | ↓ 0.1% |
Scores are final five-repeat aggregates under matched evaluation protocols. Mean-token
columns report reasoning tokens per trial; Terminal-Bench sums reasoning across agent calls.
Benchmark methodology and reproduction settings
Serving: BF16 · vLLM 0.27.1 · Qwen3 parser · context 262,144 · thinking xhigh.
Sampling: temperature 1.0 · top_p 0.95 · top_k 20 · min_p 0 · presence_penalty 0 · repetition_penalty 1.
Benchmarks: averages over five seeds (0–4) per model; five trials per task for Terminal-Bench, base and Swift 1.5 served at context 131,072 on the same Harbor build.
| Benchmark | Output cap |
|---|
| GPQA-Diamond | 100,000 |
| C-Eval | 16,384 |
| IFBench | 81,920 |
| ERQA | 100,000 |
| AIME 2026 | 250,000 |
| HMMT November 2025 | 250,000 |
| LiveCodeBench v6 | 32,768 |
| Terminal-Bench 2.1 | Agent/task limits |
## Efficiency across reasoning efforts
Qwen3.8's `reasoning_effort` setting lets users choose how much the model thinks.
For Swift 1.5 to be useful across these settings, it needs to reduce thinking while
keeping accuracy close to the base. We therefore tested `xhigh`, `medium`, and `low`:
thinking-token savings persist at every level.
| Reasoning effort |
Qwen3.8 |
Swift 1.5 |
Mean thinking reduction |
| Xhigh | 88.28% | 88.59% | ↓ 41.9% |
| Medium | 84.14% | 82.22% | ↓ 24.8% |
| Low | 84.04% | 84.85% | ↓ 28.7% |
At `low`, Swift 1.5 scores above the base while using about 29% fewer thinking tokens.
## Quantized Swift 1.5 models
| Format | Repository | Runtime |
| --- | --- | --- |
| GGUF | [Swift-1.5-Qwen3.8-27B-GGUF](https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GGUF) | llama.cpp |
| GSQ-RCO GGUF (compact 2–3 bit) | [Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF](https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF) | llama.cpp |
| AWQ INT4 (W4A16) | [Swift-1.5-Qwen3.8-27b-W4A16-AWQ](https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b-W4A16-AWQ) | vLLM (`compressed-tensors`) |
| AutoRound INT4 (W4A16) | [Swift-1.5-Qwen3.8-27b-W4A16-AutoRound](https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b-W4A16-AutoRound) | vLLM (`auto-round`) |
| AWQ + GPTQ INT4 (W4A16) | [Swift-1.5-Qwen3.8-27b-INT4](https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b-INT4) | vLLM (`compressed-tensors`) |
| NVFP4 | [Swift-1.5-Qwen3.8-27b-NVFP4](https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b-NVFP4) | NVIDIA Blackwell |
| AMD Quark FP8 (W8A8) | [Swift-1.5-Qwen3.8-27b-Quark-FP8-dynamic-AMD](https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b-Quark-FP8-dynamic-AMD) | AMD Quark |
| MLX 5-bit | [Swift-1.5-5bit-MLX](https://huggingface.co/ukisai/Swift-1.5-5bit-MLX) | Apple MLX |
| MLX 4-bit | [Swift-1.5-4bit-MLX](https://huggingface.co/ukisai/Swift-1.5-4bit-MLX) | Apple MLX |
| MLX 3-bit (text only) | [Swift-1.5-3bit-MLX-TextOnly](https://huggingface.co/ukisai/Swift-1.5-3bit-MLX-TextOnly) | Apple MLX |
These results evaluate the merged Swift 1.5 checkpoint and three INT4 exports on
**GPQA-Diamond (198 questions), IFBench (300 prompts), and AIME 2026 (30 problems)**.
Each model completed the full datasets with **one sample per prompt, seed 0, and
zero request errors**. This is a single-seed evaluation, separate from the
five-repeat BF16 release results above.
The Qwen-base columns use the **saved seed/sample 0 runs**.
Quantization recipes and serving settings differ from the new Swift 1.5 runs,
so these are reference comparisons rather than a controlled measurement of
the Swift adaptation. Token reductions below are recomputed from those same
reference samples.
| Benchmark / Swift 1.5 quantization |
Qwen base
accuracy |
Swift 1.5 quant
accuracy |
Mean token reduction |
Median token reduction |
GPQA-Diamond
AWQ INT4 | 86.36% | 88.38% | ↓ 51.5% | ↓ 64.4% |
GPQA-Diamond
AutoRound INT4 | 86.36% | 89.39% | ↓ 50.5% | ↓ 57.8% |
GPQA-Diamond
AWQ + GPTQ INT4 | 86.36% | 90.91% | ↓ 45.8% | ↓ 64.4% |
IFBench
AWQ INT4 | 72.00% | 72.00% | ↓ 36.9% | ↓ 49.3% |
IFBench
AutoRound INT4 | 72.00% | 69.33% | ↓ 29.3% | ↓ 39.2% |
IFBench
AWQ + GPTQ INT4 | 72.00% | 70.00% | ↓ 31.8% | ↓ 52.7% |
AIME 2026
AWQ INT4 | 70.00% | 86.67% | ↓ 29.2% | ↓ 36.2% |
AIME 2026
AutoRound INT4 | 76.67% | 83.33% | ↓ 17.7% | ↓ 32.4% |
AIME 2026
AWQ + GPTQ INT4 | 76.67% | 83.33% | ↓ 22.4% | ↓ 34.0% |
**AIME scoring:** truncated responses count as incorrect for both columns.
The AMD Quark INT4 and FP8 exports have separate sanity evaluations; completed
results on these three reasoning benchmarks are not available for them.
Quantized evaluation settings and BF16 reference
**Serving:** vLLM 0.29.0, tensor parallelism 1, eager execution, BF16 activations,
context 131,072, template-default thinking without an effort override. The
AWQ + GPTQ export uses FP8 KV cache; BF16, AWQ, and AutoRound use auto KV dtype.
**Sampling:** temperature 1, top-p 0.95, top-k 20, min-p 0, presence penalty 0,
repetition penalty 1, seed 0. Output caps: GPQA 100,000, IFBench 81,920,
AIME 32,768. IFBench uses official strict prompt-level scoring.
GPQA token counts cover re-tokenized reasoning; IFBench and AIME count the
full generated response. Statistics include all responses, including truncations; medians
use the midpoint of the two central values when the sample count is even.
Saved Qwen references: W4A16 for GPQA and IFBench; Qwen AWQ for the
AWQ AIME row; Qwen W4A16 for the AutoRound and AWQ + GPTQ AIME rows. The
latter is a W4A16 reference for AutoRound, not an AutoRound base run.
The new runs do not reproduce the original software stack.
The fresh Swift 1.5 BF16 reference and all quantized exports scored as follows
under this single-seed protocol:
| Model | GPQA-Diamond | IFBench strict | AIME 2026 |
| --- | ---: | ---: | ---: |
| Swift 1.5 BF16 | 91.41% | 72.00% | 86.67% |
| AWQ INT4 | 88.38% | 72.00% | 86.67% |
| AutoRound INT4 | 89.39% | 69.33% | 83.33% |
| AWQ + GPTQ INT4 | 90.91% | 70.00% | 83.33% |
Truncation counts are recorded in the linked evaluation data.
These single-seed results do not establish
quality parity or replace the broader multi-seed evaluation.
[Verified counts, token statistics, settings, and evidence hashes](https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b/blob/main/benchmarks/quantization-20260921.json).
## GGUF quantizations
| File |
Size |
KLD wikitext @512 |
KLD wikitext @32k |
99% KLD @32k |
Top-p @32k |
| Q8_0 | 29.0 GB | 0.0008 | 0.0006 | 0.005 | 98.85% |
| Q6_K_L | 25.0 GB | 0.0015 | 0.0014 | 0.010 | 98.16% |
| Q6_K | 23.9 GB | 0.0018 | 0.0016 | 0.014 | 98.30% |
| Q6_K_S | 22.9 GB | 0.0020 | 0.0016 | 0.014 | 98.24% |
| Q5_K_M | 20.9 GB | 0.0052 | 0.0061 | 0.050 | 96.92% |
| Q5_K_S | 19.6 GB | 0.0060 | 0.0069 | 0.058 | 96.91% |
| Q4_K_L | 18.8 GB | 0.0106 | 0.0103 | 0.105 | 95.79% |
| Q4_K_M | 17.4 GB | 0.0137 | 0.0134 | 0.163 | 95.03% |
| IQ4_NL | 17.4 GB | 0.0152 | 0.0140 | 0.175 | 95.39% |
| Q4_K_S | 16.4 GB | 0.0164 | 0.0154 | 0.175 | 94.84% |
| IQ4_XS | 15.5 GB | 0.0179 | 0.0173 | 0.187 | 94.96% |
| IQ3_M | 14.9 GB | 0.0410 | 0.0380 | 0.409 | 91.83% |
| Q3_K_L | 14.1 GB | 0.0442 | 0.0410 | 0.415 | 91.63% |
| Q3_K_M | 13.4 GB | 0.0570 | 0.0562 | 0.614 | 90.30% |
| IQ3_XS | 12.8 GB | 0.0583 | 0.0885 | 1.130 | 88.54% |
| Q3_K_S | 12.7 GB | 0.0648 | 0.0658 | 0.712 | 89.53% |
| IQ3_XXS | 12.3 GB | 0.0742 | 0.0844 | 0.996 | 88.70% |
| Q2_K | 10.8 GB | 0.1655 | 0.1546 | 1.728 | 84.00% |
| IQ2_M | 10.5 GB | 0.1523 | 0.1493 | 1.568 | 84.17% |
| IQ2_S | 9.7 GB | 0.2095 | 0.2589 | 3.024 | 80.58% |
| IQ2_XS | 9.1 GB | 0.2433 | 0.2622 | 2.951 | 79.99% |
| IQ2_XXS | 8.9 GB | 0.2866 | 0.2769 | 2.902 | 78.48% |
Mean KL divergence against the Swift 1.5 BF16 source, lower is better. `wikitext @512` is wikitext-2
test, 100 windows of 512 tokens. `wikitext @32k` is wikitext-2 train, 16 windows of 32,768 tokens, scoring
only the last 512 tokens of each window, so every scored token sees at least 32k tokens of context.
`99% KLD` is the 99th-percentile divergence on the same 32k run. `Top-p` is top-token agreement with BF16
on the 32k run.
Long context costs very little on this release: for every tier from `Q8_0` through `Q4_K_S`, the 32k mean
is within 10% of the 512-token value (`Q5_K_M` and `Q5_K_S` rise about 15%), and `Q4_K_M` keeps the same
top token as BF16 on 95% of positions at 32k. The pick below follows the 99th-percentile tail at 32k:
0.163 for `Q4_K_M`, 0.050 for `Q5_K_M`, 0.014 for `Q6_K`, 0.005 for `Q8_0`.
| Use case |
Pick |
| 24 GB cards, everyday use | Q4_K_M |
| Long agentic runs, strict tool-call formatting | Q6_K or higher |
| Maximum fidelity | Q8_0 |
Recipe
All 22 tiers were built with llama.cpp commit `6f41ac5` from a BF16 conversion of the published Swift 1.5
safetensors. They reuse the importance matrix and the per-tensor type layouts (`--tensor-type-file`) that
[bartowski](https://huggingface.co/bartowski/ukisai_Swift-Qwen3.8-27b-GGUF) computed for Swift 1.0 with his
[quantization-config](https://github.com/bartowski1182/quantization-config); Swift 1.5 has the same
architecture and tensor shapes. Every file was checked against BF16 on the harness above.
## License and access
Swift 1.5 is a derivative of [Qwen3.8-27B](https://huggingface.co/Qwen/Qwen3.8-27B)
(Copyright 2026 Alibaba Cloud, [Apache License 2.0](https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b/blob/main/LICENSE-APACHE-2.0)). UkisAI's contribution, including the adapted
weights, is licensed under the **[Swift Open License v1.0](https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b/blob/main/LICENSE)**. See [NOTICE](https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b/blob/main/NOTICE) for the change notice and attribution details.
Personal, research, educational, evaluation, and commercial use are free for individuals
and organizations with gross annual revenue, including affiliates, of up to US$1,000,000.
Above that threshold, commercial use requires a separate Swift Enterprise License.
Contact [UkisAI](https://ukisai.com/contact) for terms.
Nothing in the Swift Open License limits rights in Qwen3.8-27B itself under Apache 2.0.
## Citation
~~~bibtex
@misc{swift-1.5-qwen3.8-27b,
title = {Swift 1.5 Qwen3.8-27B},
author = {UkisAI},
year = {2026},
url = {https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b}
}
~~~
## Acknowledgements
We acknowledge the [NVIDIA Innovation Lab](https://www.nvidia.com/en-us/data-center/innovation-lab/),
[Amazon Web Services](https://aws.amazon.com/), and
[Google Cloud](https://cloud.google.com/) for providing compute credits and
infrastructure support for Swift's development, training, and evaluation.
