@sgl_project: We turned Qwen3.8-27B into a multimodal decision model. It beat Pokémon FireRed’s elite four and champion with sub-100 …
Summary
SGLang project turned Qwen3.8-27B into a multimodal decision model that beat Pokémon FireRed's elite four and champion with sub-100 ms decisions from live game state, and shipped native /v1/decisions plus /v1/systemone endpoints for classifying, scoring, and integrating open models with the TypeSafe SDK.
View Cached Full Text
Cached at: 10/01/26, 04:15 AM
We turned Qwen3.8-27B into a multimodal decision model.
It beat Pokémon FireRed’s elite four and champion with sub-100 ms decisions from live game state.
With SGLang’s native /v1/decisions, you can now turn LLMs and VLMs into classification and scoring models.
We also added /v1/systemone so Jev-like open models can work with the TypeSafe SDK.
Similar Articles
@sgl_project: The king of small models is back! Qwen3.8-27B from @Alibaba_Qwen is open source, and Day-0 support is live in SGLang: -…
Qwen3.8-27B, a 27B-parameter multimodal AI model from Alibaba, is now open source with day-0 support in SGLang, offering high inference speeds and superior performance in coding and office tasks.
@HuggingModels: Ever wanted a model that combines Qwen's reasoning, Opus's creativity, and GLM's efficiency? Meet Qwen-3.5-Opus-GLM-27B…
Qwen-3.5-Opus-GLM-27B is a 27B parameter GGUF merge combining the strengths of Qwen's reasoning, Opus's creativity, and GLM's efficiency, designed for high-performance local AI without cloud dependency.
@sudoingX: update: qwen 3.6 27b dense q4 just one shotted octopus invaders game on a single 3090. hermes agent drove the whole thi…
A user benchmark demonstrates that the Qwen 3.6 27B dense model (Q4 quantized) can autonomously generate a fully playable multi-file game in a single prompt on a single RTX 3090, significantly outperforming its predecessor with zero manual interventions. The results highlight major improvements in local code generation and agentic capabilities for consumer-grade hardware.
Perplexity Decider 27B: Open weights decision model fine tune of Qwen3.8 27B
Perplexity has released pplx-decider-v1-27b, an open-weights decision model fine-tuned from Qwen3-27B that scores 85.71% overall across 11 benchmarks, notably improving on tasks like RAGTruth (88.80%), FinancialPhraseBank (84.18%), and TabFact (90.60%). The model can be run locally via Hugging Face and the Perplexity API for structured choice and yes/no probability predictions.
I just ran Qwen 3.8 27 in Q4 against GPT 5.6 Sol high - and it easily won against SOL - complex animated SVG tasks
A user tested Qwen 3.8 27B in Q4 quantization against GPT 5.6 SOL in complex animated SVG tasks and found that Qwen performed better with fewer errors, highlighting strengths in spatial reasoning and coding.