BigBang-v1 vs Hy3

BigBang-v1

A Qwen3.6-35B-A3B fine-tuned on synthetic research problems, under plain Apache 2.0, for search, science and coding work. The publisher ships BF16 only, about 86 GB to run, and has no report, safety results or replies to questions, so smaller builds and support come from others.

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.

BigBang-v1

1. On the reference machine use vllm serve endless-frontier/BigBang-v1 --max-model-len 131072 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder
2. The card says to keep at least 128K of context so its thinking isn't cut short
3. Thinking control isn't documented, try the base template's enable_thinking false in chat_template_kwargs and check the output
4. For a smaller build you'll need someone else's quantisation, the publisher has none
5. Validate JSON output yourself or use the runtime's guided decoding

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