Model card written by GPT-6 Astra, refined with human feedback:

Swift 1.5 HyperQwen — INT4 heads

Swift 1.5's efficient reasoning, combined with HyperQwen's optimized serving, delivered 37% lower average task completion time than HyperQwen serving the Qwen fast checkpoint in our RTX 3090 evaluation.

Browse all three Swift HyperQwen variants.

Changes from upstream

Upstream checkpoint: ukisai/Swift-1.5-Qwen3.8-27b-W4A16-AWQ, revision 9dba8a05. The changes below are relative to this AWQ checkpoint.

Preserve the upstream AWQ INT4 model body.
Convert embeddings to INT8.
Quantize the main output head (lm_head) and eight MTP linear matrices to GPTQ INT4, directly from upstream BF16 tensors.
Rebuild the MTP draft shortlist using Swift responses: 65,536 tokens, while retaining the target model's full vocabulary.
No additional fine-tuning; Swift's reasoning-efficiency training is retained.

Performance

Measurement	Qwen — HyperQwen fast quant (W4A16 AutoRound)	Swift 1.0	Swift 1.5 INT8 heads	Swift 1.5 INT4 heads
Average request time ↓	108.1 s	66.2 s	72.2 s	68.2 s
Average output tokens/task ↓	8,985	5,245	5,751	5,669
Median decode tokens/s ↑	112.1	105.9	104.0	107.2
Total output tokens, 630 tasks	5.66M	3.30M	3.62M	3.57M

Quality

Test	Qwen — HyperQwen fast quant (W4A16 AutoRound)	Swift 1.0	Swift 1.5 INT8 heads	Swift 1.5 INT4 heads
GSM8K — 200-question subset	97.5%	98.0%	98.0%	97.5%
IFBench — 300 prompts, strict	74.0%	73.3%	73.7%	72.3%
LiveCodeBench — 100-problem subset	90%	89%	89%	91%
Custom tool-call/JSON checks — 30 tasks	29/30	28/30	30/30	30/30
Perplexity, English/Python ↓	6.551	6.605	6.643	6.679
Truncated answers, counted wrong	2	1	1	0

GSM8K: the first 200 questions from the test split, with thinking disabled.
IFBench: all 300 prompts in the pinned IFBench test dataset, scored with the official strict prompt-level verifier.
LiveCodeBench: a frozen v6-era dataset subset of Python stdin/stdout problems: 34 easy, 33 medium, 33 hard. Scored against supplied public/private tests with a custom judge; not a full official LiveCodeBench result.
Tool-call/JSON checks: 20 custom weather-tool tasks checking the function name, arguments, Celsius-to-Fahrenheit conversion and final JSON; plus 10 JSON inventory-filtering tasks. These are integration checks, not an external agent benchmark.
Perplexity: 18,729 scored tokens from English Wikipedia and Python source; lower is better. Recomputed from saved token log-probabilities, excluding Danish; see subset results.

Compared with the Swift 1.5 INT8-head variant, this INT4-head conversion measured 3.2% higher decode TPS and 5.5% lower average request time; benchmark score changes were mixed.

Evaluation setup: RTX 3090 24 GB; FP8 KV cache; 150,000-token configured context; 128,000 output tokens per call. These are runtime settings, not fixed model properties, and the TPS test uses short prompts. Task evaluation used two concurrent requests; TPS was measured with one request at a time. All models used the same serving settings and task budgets. Thinking tests used xhigh effort, temperature 1.0, top_p 0.95, top_k 20 and seed 15027; GSM8K/tool checks were greedy.

Setup

Requires the patched HyperQwen runtime, not stock vLLM or GGUF tools. From an installed HyperQwen checkout:

hf download daavidhauser/Swift-1.5-Qwen3.8-27B-W4A16-HyperQwen-INT4 --local-dir models/Swift-1.5-Qwen3.8-27B-W4A16-HyperQwen-INT4
MODEL="$PWD/models/Swift-1.5-Qwen3.8-27B-W4A16-HyperQwen-INT4" CTX=long MAX_LEN=150000 SPEC=mtp \
  bash single-user/start_qwen.sh

See RUNTIME.md for the pinned runtime, launcher snapshot, complete evaluated settings and installation notes. The checkpoint is already converted: do not requantize its heads. Vision weights are retained, but the evaluation is text-only. Multi-user batch settings were not benchmarked in this campaign.

Detailed results and evaluation code are in evaluation/. Quantization/source provenance is included with the model. The upstream Swift Open License v1.0, Apache 2.0 base-model license, and NOTICE are retained.

The upstream AWQ quantization is UkisAI's Swift 1.5 W4A16-AWQ. Swift's training is by UkisAI; the serving runtime is HyperQwen. This is an independent conversion with local evaluation by daavidhauser.