Qwen

Qwen3-8B 100 clients

5

Qwen 2.5 3B

5

Qwen2.5-1.5B

5

Qwen3-235B-A22B-Thinking-2507 generated text business medical legal (test)

5

Qwen models on RTX 3090 (24GB VRAM) 2.5 (train)

5

Qwen models on A6000 (48GB VRAM) 2.5 (train)

5

Qwen2.5-14B (2048 samples)

5

Qwen3-32B 16×H100s

5

Qwen held-out (test)

5

Qwen VL 235B 3

5

Qwen-3-VL-30B

5

Qwen2.5-7B 64K context (train)

5

Qwen-2

5

Qwen2.5-3B Layer 24 activations (held-out)

4

Qwen3-8B layer 20

4

Qwen3-14B S=2048 (base variant)

4

Qwen3-32B

4

Qwen 14B Instruct 2.5

4

Qwen2.5-VL-7B

4

Qwen3-235B-A22B

4

Qwen3-30B-A3B-Instruct-2507 256K context window

4

Qwen3-30B-A3B-Instruct-2507 128K context window

4

Qwen3-30B-A3B-Instruct-2507 64K context window

4

Qwen3-30B-A3B-Instruct-2507 32K context window

4

Qwen (unseen)

4