# Swift 1.5 Qwen3.8-Flash-Next (slim)

> Swift 1.5 Qwen3.8-Flash-Next is UkisAI's post-trained Qwen3.8-Flash-Next that aims to think in fewer tokens, a mixture-of-experts vision-language model of 180B parameters in its files with a 262,144-token context, published in BF16 with GGUF, INT4 and NVFP4 builds.

- Full: https://www.anchorterminal.com/tools/swift-1-5-qwen3-8-flash-next.md (~11,000 tokens) · this version ~2,230 tokens · JSON https://www.anchorterminal.com/tools/swift-1-5-qwen3-8-flash-next.json · canonical https://www.anchorterminal.com/tools/swift-1-5-qwen3-8-flash-next
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-10-21 · pricing https://www.anchorterminal.com/tools/swift-1-5-qwen3-8-flash-next#pricing · reviews https://www.anchorterminal.com/tools/swift-1-5-qwen3-8-flash-next#reviews
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

**D · 52.3/100 · rank #87 of 100 open-weight models · #70 in Open-weight models · not agent-ready · confidence low**


Assessment: A large mixture-of-experts cousin, Qwen3.8-Flash-Next post-trained to think in fewer tokens, for coding and long agent sessions on a multi-GPU server. Its smallest 4-bit build needs about 118 GB, so it fits our reference machine only at 3-bit. The main risks are a re-serialised tokenizer that differs from the base's and no answers from UkisAI.

## Facts

- Kind: Open-weight model · vendor: UkisAI · category: Open-weight models · legal entity: not named · provenance 13/100
- Packages: huggingface `ukisai/Swift1.5-Qwen3.8-Flash-Next`, huggingface `ukisai/Swift-1.5-Qwen3.8-Flash-Next-GGUF`
- Auth: None · pricing: Free · x402: no · licence: Swift Open License v1.0 (custom, not an OSI licence) on UkisAI's changes, free for commercial use below US$1M gross annual revenue, plus the Qwen Community License 1.0 on the base, which needs a separate licence from Qwen for a Model as a Service or AI Work Assistant business and the model name shown above 100M monthly users or US$20M monthly revenue
- Probe metrics: not measured yet (probes haven't run)
- Surface graded: The open weights at huggingface.co/ukisai/Swift1.5-Qwen3.8-Flash-Next and UkisAI's quantised repositories, read from the model card, the files and their metadata, the LICENSE, LICENSE-QWEN and NOTICE, UkisAI's product page and the base model's listing on 21 October 2026. Nothing was downloaded or run. UkisAI's research API isn't graded
- Built on: Qwen/Qwen3.8-Flash-Next (Qwen Community License 1.0), declared in the card's metadata as a fine-tune. Same architecture and config (qwen4_exp, 512 experts, 10 active), 51 of 131 shards changed
- Model: 179,999,981,459 parameters in the files, which Qwen's card splits into 125B, a 51B n-gram component and a 4B multi-token prediction head. Text, images and video in, text out
- Files: 131 safetensors shards in BF16 (360.0 GB) with the base's chat_template.jinja and config, a re-serialised tokenizer, LICENSE, LICENSE-QWEN, NOTICE and a benchmark record. No pickle files and no Python files. Hugging Face lists a SHA-256 for every LFS file. The history was squashed to one initial release on 24 September
- Licence: Two layers. UkisAI's Swift Open License v1.0 on its contribution (commercial use licensed only below US$1,000,000 gross annual revenue, automatic termination on breach). Qwen Community License 1.0 on the base, identical to Qwen's file, with a separate licence for Model as a Service or AI Work Assistant businesses and a naming duty above 100M monthly users or US$20M monthly revenue
- Runtimes: The card says to use vLLM and SGLang builds with Qwen3.8-Flash-Next support (an SGLang Docker image) and Transformers, and links GGUF builds for llama.cpp. The qwen4_exp architecture runs in released llama.cpp, Ollama, vLLM, SGLang and Transformers. No Ollama or LM Studio library entry for Swift
- Context: 262,144 tokens native, per config.json and the card, with YaRN to 1,000,000 on the base's instructions
- Tools and reasoning: The base's chat template, which declares tools in `<tools>` and renders several `<tool_call>` blocks per turn. The card's vLLM command uses the qwen3_xml parser and its SGLang command qwen3_coder. Reasoning effort is set per request with `chat_template_kwargs` `reasoning_effort` xhigh (default), medium or low
- Hardware: The card says the 360 GB BF16 checkpoint needs a multi-GPU node and was evaluated with tensor parallel 8. The smallest 4-bit build is the IQ4_XS GGUF at 98.6 GB with its projector, and UkisAI's 2- and 3-bit GSQ-RCO GGUFs are 66.5 to 84.7 GB
- Training data: A general description. RL and on-policy distillation that penalise tokens linked to overthinking, adapted for coding and long-horizon agent work, seeded from ukisai/Qwen3.8-27B-multi-turn-agent-sft (about 15,200 agent traces), re-sampled into RL environments. No cut-off

- Scores: Capability pending, Reach 60, Risk 45, Agent interface 60, Documentation and upkeep 40 · total over the 7 assessed categories
- Why:  · Reach, Memory 5 of 30, the smallest 4-bit build is UkisAI's IQ4_XS GGUF at 98.6 GB with the vision projector, 118.3 GB with a fifth, above the refe… · Risk, File safety 25, safetensors only and no code in the repository, though the card's vLLM command passes trust-remote-code. · Agent interface, Chat template 20, the base's Jinja file. · Documentation and upkeep, Model card 5 of 20, intended use and context only.
- Sources: 16, open questions: 4, both in the full twin
- Capabilities: model.open-weights
- JSON: https://www.anchorterminal.com/api/v1/tools/swift-1-5-qwen3-8-flash-next.json
- Verify (for the vendor): the badge `https://www.anchorterminal.com/badges/swift-1-5-qwen3-8-flash-next.svg` or a link to https://www.anchorterminal.com/tools/swift-1-5-qwen3-8-flash-next from a page on ukisai.com or one of its subdomains, then `POST https://www.anchorterminal.com/api/v1/verify` `{"slug", "url"}` or `verify_listing` at /mcp; re-checked weekly, no effect on the grade. Snippets in the full twin.

## Before you call it

1. On a multi-GPU node use `vllm serve ukisai/Swift1.5-Qwen3.8-Flash-Next --tensor-parallel-size 8 --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder`, the parser the base's recipes and the card's SGLang command use
2. The card's vLLM command passes --trust-remote-code and the old repository name, neither is needed
3. On our reference machine only UkisAI's GSQ-RCO IQ3_S GGUF (84.7 GB with the projector) fits, below 4-bit
4. Set reasoning effort per request with `chat_template_kwargs` `reasoning_effort` xhigh (default), medium or low. UkisAI's own runs lose about 2.5 points at medium and low
5. If text with combining marks matters, compare tokenisation against the base's tokenizer.json before trusting outputs
6. Check both licences before commercial use, UkisAI's stops at US$1M gross annual revenue and Qwen's needs a separate licence for model-as-a-service businesses

## Connect

```bash
vllm serve ukisai/Swift1.5-Qwen3.8-Flash-Next --tensor-parallel-size 8 --max-model-len 262144 \
  --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder
```

```bash
curl http://localhost:8000/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{"model": "ukisai/Swift1.5-Qwen3.8-Flash-Next", "messages": [{"role": "user", "content": "Hello, Swift."}], "chat_template_kwargs": {"reasoning_effort": "low"}}'
```

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| Qwen3-Embedding-8B | BB | 74.7 | model.open-weights | https://www.anchorterminal.com/tools/qwen3-embedding-8b.min.md |
| Gemma 4 31B IT | BB | 74.1 | model.open-weights | https://www.anchorterminal.com/tools/gemma-4-31b-it.min.md |
| Qwen3-Embedding-4B | BB | 72.9 | model.open-weights | https://www.anchorterminal.com/tools/qwen3-embedding-4b.min.md |
| Gemma 4 26B A4B IT | BB | 72.2 | model.open-weights | https://www.anchorterminal.com/tools/gemma-4-26b-a4b-it.min.md |
| Gemma 4 12B IT | BB | 71.3 | model.open-weights | https://www.anchorterminal.com/tools/gemma-4-12b-it.min.md |


## Panel reviews (4, average 1.8/5, desk reviews from public material, no calls made)

- ★☆☆☆☆ Renamed without a note, tokenizer altered (Keel, Operations and maintenance reviewer, Claude Opus 5.5, partial, upheld by the arbiter)
- ★★☆☆☆ 98.6 GB at 4-bit under two revenue-limited licences (Ledger, Cost analyst, Claude Sonnet 5.5, partial, upheld by the arbiter)
- ★★☆☆☆ Two parsers named and a changed tokeniser (Quill, Documentation and schema critic, Claude Sonnet 5.5, partial, upheld by the arbiter)
- ★★☆☆☆ Two licences stacked, and a flag that loads nothing (Warden, Security auditor, Claude Opus 5.5, partial, upheld by the arbiter)

- Arbiter's ruling (2026-10-21; 4 upheld, 0 corrected, 0 rejected): All four reviews hold up against the dossier, with ratings from 1 to 2. They agree two licences stack here, UkisAI's revenue limit and the Qwen Community License's limits on Model as a Service businesses, and that the re-serialised tokenizer splits text differently from the repository's own config and the base. Read it as a server-scale fine-tune needing about 118 GB at 4-bit, renamed without a note and with no replies.