Qwen
Two Qwen2.5 text models and seven Qwen3.x vision-language models in the release matrix.
Qwen is the widest family in the matrix and the one that covers the most architectural ground. Qwen2.5 is a pair of conventional dense text decoders, one of them specialised for code. Qwen3.x is vision-language, and between its seven entries it spans dense decoders, sparse mixtures of experts, Gated DeltaNet, and — in Qwen3.8 Flash-Next — Qwen Sparse Attention with an n-gram embedding table.
That range is why this page is worth reading rather than skimming: none of those shapes needed an operator library of its own. They compile through the same pipeline as everything else on the overview, and the benchmark snapshot keeps their differences in separate rows rather than turning them into separate install paths.
Qwen2.5 14B Instruct
- Kind: Language model
- Task: Text generation
- Parameters: 14B
- Architecture: Dense transformer
- Quantization: int4 RTN, group 128
- Base model: Qwen/Qwen2.5-14B-Instruct
- Validation status: Validated
- Measured in v0.5.3: 26.81 tokens/s, 1.98 s to first token, at a 2048-token prompt — all three lengths
Qwen2.5-Coder 14B Instruct
A coding assistant that fits in local memory and holds a 16K-token file in its context without falling over.
- Kind: Language model
- Task: Code generation
- Parameters: 14B
- Architecture: Dense transformer
- Quantization: int4 RTN, group 128
- Base model: Qwen/Qwen2.5-Coder-14B-Instruct
- Validation status: Validated
- Measured in v0.5.3: 26.80 tokens/s, 1.98 s to first token, at a 2048-token prompt — all three lengths
Qwen3.5 9B
- Kind: Vision-language model
- Task: Image understanding, text generation
- Parameters: 9B
- Architecture: Dense transformer + vision encoder
- Quantization: int4 text decoder, group 32; fp16 vision encoder
- Base model: Qwen/Qwen3.5-9B
- Validation status: Validated
- Measured in v0.5.3: 41.26 tokens/s, 4.13 s to first token, at a 2048-token prompt — all three lengths
Qwen3.5 35B-A3B
- Kind: Vision-language model
- Task: Image understanding, text generation
- Parameters: 35B total, 3B active
- Architecture: Sparse mixture of experts + vision encoder
- Quantization: int4 text decoder, group 32; fp16 vision encoder
- Base model: Qwen/Qwen3.5-35B-A3B
- Validation status: Validated
- Measured in v0.5.3: 68.40 tokens/s, 3.07 s to first token, at a 2048-token prompt — all three lengths
Qwen3.5 122B-A10B
- Kind: Vision-language model
- Task: Image understanding, text generation
- Parameters: 122B total, 10B active
- Architecture: Sparse mixture of experts + vision encoder
- Quantization: int4 text decoder, group 32; fp16 vision encoder
- Base model: Qwen/Qwen3.5-122B-A10B
- Validation status: New in v0.5.3
- Measured in v0.5.3: 23.93 tokens/s, 9.33 s to first token, at a 2048-token prompt — all three lengths
Qwen3.6 27B
- Kind: Vision-language model
- Task: Image understanding, text generation
- Parameters: 27B
- Architecture: Dense transformer + vision encoder
- Quantization: int4 RTN text decoder, group 32; fp16 vision encoder
- Base model: Qwen/Qwen3.6-27B
- Validation status: Validated
- Measured in v0.5.3: 13.86 tokens/s, 10.50 s to first token, at a 2048-token prompt — all three lengths
Qwen3.6 35B-A3B
Gated DeltaNet and a sparse MoE in the same graph. Nothing about this model is a stock transformer, and it still compiles to one artifact.
- Kind: Vision-language model
- Task: Image understanding, text generation
- Parameters: 35B total, 3B active
- Architecture: Gated DeltaNet + sparse mixture of experts
- Quantization: int4 text decoder, group 32; fp16 vision encoder
- Base model: Qwen/Qwen3.6-35B-A3B
- Validation status: Validated
- Measured in v0.5.3: 68.38 tokens/s, 3.07 s to first token, at a 2048-token prompt — all three lengths
Qwen3.8 27B
Added in v0.4.0. A model that did not exist when the compiler was designed, brought up without a new operator library.
- Kind: Vision-language model
- Task: Image understanding, text generation
- Parameters: 27B
- Architecture: Dense transformer + vision encoder
- Quantization: int4 k-quant text decoder, group 128; fp16 vision encoder
- Base model: Qwen/Qwen3.8-27B
- Validation status: Validated
- Measured in v0.5.3: 14.76 tokens/s, 9.33 s to first token, at a 2048-token prompt — all three lengths
Qwen3.8 Flash-Next
Added in v0.5.3. Hybrid attention, a sparse MoE, and an off-accelerator n-gram embedding table in one compiled graph.
- Kind: Vision-language model
- Task: Image understanding, text generation
- Parameters: 125B total, 6B active
- Architecture: Hybrid Gated DeltaNet + QSA + sparse MoE + n-gram embedding
- Quantization: int4 RTN text decoder, group 32; fp8 n-gram embedding; fp16 vision encoder
- Base model: Qwen/Qwen3.8-Flash-Next
- Validation status: New in v0.5.3
- Measured in v0.5.3: 31.16 tokens/s, 11.42 s to first token, at a 2048-token prompt — all three lengths