{"slug":"ds4-vision-2x6000-jj-dspark-k3","title":"DeepSeek V4 Flash Vision on 2x RTX PRO 6000 Blackwell \\x{2014} TP2 DSpark K3","summary":"Community image TP2 DSpark K3 Vision serving recipe: B12X sparse MoE, FP8 KV, three draft layers, 1M context. Saved numbers from the KK comparison.","license":"CC-BY-4.0","tier":"community","beta":false,"revision":{"no":2,"commit_sha":"17fd5669c2c9eaeb55ffb611f348686755997213","content_hash":"a67645b428d443142e2632eb2d7b7ff2eb8c27c4","created_at":"2026-10-20 21:54:52.636866-06","notes":"Vision compose: MODE vision profile, request slots 4, graph cap 16, speculative depth > 3 unsupported."},"engine_source":{"type":"docker_image","git_url":"https://github.com/voipmonitor/vllm/tree/integration/jovian-reviewed-sources-20260911","base_engine":"vllm","pip_package":null,"base_version":"jovian-judgement r9 (vllm f66599d / b12x 15b6813)","image_digest":"voipmonitor/vllm:jovian-judgement-vllmf66599d-b12x15b6813-fi803c466-cu133-torch213-20260907-r9","install_note":null,"pinned_commit":null},"credit":[{"name":"local-inference-lab","url":"https://github.com/local-inference-lab/rtx6kpro"}],"author":{"name":"localai-recipes","isSite":true},"uploaded_by":"localai-recipes","config":{"launch":["docker","compose","-f","docker-compose-ds4-vision-jovian-judgement-r9.yml","up","-d"]},"models":[{"slug":"deepseek-v4-flash-vision-exp","name":"DeepSeek-V4-Flash-Vision-Exp","hf_repo":"deepseek-ai/DeepSeek-V4-Flash-Vision-Exp","quant":"fp8"}],"hardware":[{"slug":"nvidia-rtx-pro-6000-blackwell-max-qx2","vendor":"nvidia","card":"RTX PRO 6000 Blackwell Max-Q","card_family":"rtx-pro-6000-blackwell","vram_gb":96,"device_count":2}],"generation":"blackwell","tags":["dspark","jovian-judgement","vision"],"tasks":["chat","code","vision"],"aa":{"aa_index":35,"aa_provenance":"published","aa_source_url":"https://artificialanalysis.ai/models/deepseek-v4-flash-vision","aa_index_version":null,"aa_methodology_version":"v4.3.2","aa_published_fetched_at":"2026-10-06 18:30:00-06","aa_family_slug":"deepseek-v4-flash-vision-exp","aa_resolved_slug":"deepseek-v4-flash-vision-exp"},"performance":[{"claimed_by":"submitter","metrics":"{\"points\": [{\"notes\": \"Saved JJ (KK comparison), 5-window median, ctx-0 decode\", \"concurrency\": 1, \"context_len\": 1, \"decode_tok_s\": 169.19999999999998863131622783839702606201171875}, {\"notes\": \"Saved JJ, aggregate C4 (request slots = 4)\", \"concurrency\": 4, \"context_len\": 1, \"decode_tok_s\": 394.69999999999998863131622783839702606201171875}, {\"notes\": \"Saved JJ, uncached nominal-32K prefill\", \"concurrency\": 1, \"context_len\": 32768, \"prefill_tok_s\": 8993}]}"}],"perf_points":[{"id":1738,"context_len":1,"prefill_tok_s":null,"decode_tok_s":169.19999999999998863131622783839702606201171875,"ttft_ms":null,"concurrency":1,"claimed_by":"submitter","notes":"Saved JJ (KK comparison), 5-window median, ctx-0 decode"},{"id":1739,"context_len":1,"prefill_tok_s":null,"decode_tok_s":394.69999999999998863131622783839702606201171875,"ttft_ms":null,"concurrency":4,"claimed_by":"submitter","notes":"Saved JJ, aggregate C4 (request slots = 4)"},{"id":1740,"context_len":32768,"prefill_tok_s":8993,"decode_tok_s":null,"ttft_ms":null,"concurrency":1,"claimed_by":"submitter","notes":"Saved JJ, uncached nominal-32K prefill"}],"agent":{"prompt":"Fetch https://localairecipes.com/api/v1/recipes/ds4-vision-2x6000-jj-dspark-k3.json and apply the inference recipe it describes (engine flags, model, quant, context) to my setup. Ask before overwriting any existing config.","curl":"curl -fsSL https://localairecipes.com/api/v1/recipes/ds4-vision-2x6000-jj-dspark-k3.json"},"requirements":{"requires":["Docker (or a compatible OCI container runtime)","NVIDIA Container Toolkit (nvidia-container-toolkit) - exposes host GPUs to the container"],"driver_floor":{"cuda":"13.3","linux_min":">=580","text":"NVIDIA driver >= 580 (image CUDA 13.3 - CUDA 13.x applications run on drivers >= 580 under CUDA minor-version compatibility; the CUDA 13.3 toolkit shipped driver >= 610.43.02)","source":"NVIDIA CUDA Toolkit Release Notes - driver compatibility tables (https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/, fetched 2026-10-03)"},"placeholders":["Hardware-bound values in this recipe (device counts, bind-mount paths, utilization, ports) describe the author's machine, not yours - map them to your own rig before launching. Keep the engine flags, model, quant, and parallelism strategy; adjust paths and ports to the hardware you actually have.","GPU identity: list your GPUs with `nvidia-smi -L` and substitute your own GPU UUIDs/indices wherever the recipe pins the author's.","Model directory: bind-mount paths are author-side - point them at your own mount where the model files live.","Host port: pick a free port on your machine and publish the container there; keep the in-container port unchanged."]},"artifacts":[],"links":{"self":"/api/v1/recipes/ds4-vision-2x6000-jj-dspark-k3.json","alternate":"/r/ds4-vision-2x6000-jj-dspark-k3","describedby":"/openapi.json"},"updated_at":"2026-10-20 21:54:52.636866-06","created_at":"2026-10-20 21:49:12.715911-06","speeds":{"decode_tok_s":169.19999999999998863131622783839702606201171875,"prefill_tok_s":8993,"leaderboard":"/api/v1/leaderboard.json"}}