Darwin-36B-Opus vs Hy3

Darwin-36B-Opus

A text-only reasoning cousin merged from Qwen3.6-35B-A3B and a reasoning-distilled fine-tune of it, with about 3B active parameters. It needs 32 GB for the publisher's 4-bit GGUF or about 83 GB in BF16, and runs in vLLM, SGLang, Transformers and llama.cpp. The main risks are an undisclosed merge method, a card that disagrees with its files, and no safety results.

Hy3

Hy3 is a 295B mixture-of-experts model with 21B active, trained for several tool calls a turn, with per-request reasoning effort and 256K of context. It needs an 8-GPU server for FP8 or about 200 GB for a 4-bit build. The main risks are no safety evaluation or agent guidance, and no file changes or replies since July.

Darwin-36B-Opus

1. Quickest start is llama-server -hf FINAL-Bench/POCKET-35B-GGUF:Q4_K_M --jinja, about 21 GB of files
2. On the reference machine BF16 fits, vllm serve FINAL-Bench/Darwin-36B-Opus --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder, the parsers for Qwen3.6's format
3. Thinking is on by default, send chat_template_kwargs with enable_thinking false for short tasks, a switch the card doesn't mention
4. The card suggests temperature 0.6 to 0.7 and at least 5,120 new tokens for its think traces
5. generation_config.json stops only on the end-of-text token, so add <|im_end|> as a stop token if a runtime keeps generating after the turn ends
6. Text only, the vision encoder of Qwen3.6-35B-A3B isn't in these files
7. Validate JSON output yourself, since structured output isn't documented

Hy3

1. Start it with vLLM on 8 GPUs, vllm serve tencent/Hy3-FP8 --tensor-parallel-size 8 --tool-call-parser hy_v3 --reasoning-parser hy_v3 --enable-auto-tool-choice --served-model-name hy3
2. In SGLang use --tool-call-parser hunyuan --reasoning-parser hunyuan, as the card shows
3. Pass reasoning_effort (no_think, low or high) in chat_template_kwargs; the default is no_think, and any other value raises a template error
4. With the vLLM reasoning parser the chain of thought comes back in message.reasoning and the answer in message.content
5. Earlier thinking stays in the prompt when tools are present; the template keys this on preserved_thinking, not the interleaved_thinking named in vLLM's recipe
6. Use temperature 0.9 and top_p 1.0, the card's recommended sampling
7. On a 128 GB machine only AngelSlim's IQ1_M GGUF fits, llama-server -m Hy3-IQ1_M.gguf -ctk q8_0 -ctv q8_0 -fa on -c 65536, with llama.cpp b11768 or later; at 89.4 GB × 1.2 it sits right at the 107.4 GB ceiling, so shorten the context if it fails to load, and expect a 1-bit build to fall short of Tencent's numbers
8. Without the hardware, call it through Featherless AI or DeepInfra via Hugging Face's inference providers