DeepSeek-V4-Flash-0731 vs Qwen3.6-35B-A3B

DeepSeek-V4-Flash-0731

A long-context agent cousin under MIT with documented tool calls, three reasoning effort levels and support in every major runtime but Ollama. It needs about 190 to 200 GB of memory, though six inference providers serve it. The main risks are prompts built from DeepSeek's Python, no published safety results and no answers in its repository.

Qwen3.6-35B-A3B

A fast mixture-of-experts cousin with 3B active parameters for coding and tool calling, with image and video input, in every major runtime under plain Apache 2.0. Ollama's 4-bit build needs about 28 GB. The main risk is that nobody at Alibaba maintains it, with no safety results, no security contact and no replies to tool-calling reports.

DeepSeek-V4-Flash-0731

1. vllm serve deepseek-ai/DeepSeek-V4-Flash-0731 --trust-remote-code --kv-cache-dtype fp8 --block-size 256 --data-parallel-size 4 --enable-expert-parallel --moe-backend deep_gemm_mega_moe --attention-config '{"use_fp4_indexer_cache": true}' --speculative-config '{"method":"dspark","num_speculative_tokens":7,"draft_sample_method":"greedy"}' --tokenizer-mode deepseek_v4 --tool-call-parser deepseek_v4 --enable-auto-tool-choice --reasoning-parser deepseek_v4 (the card's 4x GB300 example plus the parser flags from vLLM's recipe)
2. Build prompts with encoding_dsv4.py, deepseek-recipe's V4 encoder or vLLM's --tokenizer-mode deepseek_v4, since the repository has no Jinja template
3. Tool calls come back in a DSML tool_calls block with invoke elements, string parameters carry string="true" and JSON values string="false"
4. Set temperature 1.0 and top_p 0.95 for agent work, and allow up to 384K output tokens at reasoning effort high or max
5. Without the hardware, call it through a Hugging Face inference provider (Together, DeepInfra, Novita, Baseten, Scaleway, Featherless), not DeepSeek's API, where deepseek-v4-flash now means V4.1 Flash
6. LM Studio's library build is a 156.4 GB MXFP4 GGUF and LM Studio states 156 GB minimum system memory

Qwen3.6-35B-A3B

1. Start it with ollama run qwen3.6:35b (Q4_K_M, 23 to 24 GB on disk, about 28 GB in memory)
2. On the reference machine use vllm serve Qwen/Qwen3.6-35B-A3B --max-model-len 131072 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder, BF16, since vLLM doesn't document FP8 compute on gfx1151
3. Use the qwen3_coder tool-call parser and the qwen3 reasoning parser in vLLM 0.19.0 or SGLang 0.5.10 or later
4. Thinking is on by default and /nothink isn't supported, so pass enable_thinking as false in chat_template_kwargs for short tasks, or preserve_thinking as true to keep earlier reasoning
5. Card sampling for thinking is temperature 1.0, top_p 0.95, top_k 20, presence_penalty 1.5, and for non-thinking temperature 0.7, top_p 0.8
6. Keep at least 128K of context as the card advises, and turn on YaRN only beyond 262,144 tokens
7. Validate JSON output yourself or use the runtime's guided decoding, since the model card documents none