gpt-oss-20b vs Hy3

gpt-oss-20b

gpt-oss-20b is a cousin for reasoning, coding and single tool calls with structured output, under plain Apache 2.0. It needs about 17 GB with a working context, and five inference providers serve it. The main risks are that it trails o4-mini at keeping system instructions above user ones, and OpenAI hasn't answered a report since August 2025.

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.

gpt-oss-20b

1. Start it with ollama run gpt-oss:20b (the build that fits the reference machine), or vllm serve openai/gpt-oss-20b with a current vLLM release rather than the card's pinned vllm==0.10.1+gptoss wheel
2. Use the harmony format through the chat template or openai-harmony; OpenAI says the model doesn't work correctly without it
3. Parse tool calls and reasoning with openai-harmony; calls come on the commentary channel ending in <|call|>, reasoning on the analysis channel
4. Set effort with Reasoning: high (or low, medium) in the system message, or reasoning_effort in the template; medium is the default
5. Ask for JSON by putting a JSON Schema under # Response Formats in the developer message
6. Drop earlier reasoning after a final answer but pass it back after tool calls, and don't show the analysis channel to users
7. Expect one tool call per turn; run several in sequence rather than in parallel

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