From 0f18d389b46bdb8d8860481019f55b402ec65a06 Mon Sep 17 00:00:00 2001 From: zijiexia <37504505+zijiexia@users.noreply.github.com> Date: Tue, 1 Sep 2026 13:09:14 -0700 Subject: [PATCH] [Cookbook] Verify DeepSeek-V4 Flash Vision balanced and high-throughput on B200 (#37468) Co-authored-by: Claude Fable 5 --- .../configs/deepseek-ai/deepseek-v4-benchmarks.jsx | 14 ++++++++++++-- .../snippets/configs/deepseek-ai/deepseek-v4.jsx | 6 ++---- 2 files changed, 14 insertions(+), 6 deletions(-) diff --git a/docs/src/snippets/configs/deepseek-ai/deepseek-v4-benchmarks.jsx b/docs/src/snippets/configs/deepseek-ai/deepseek-v4-benchmarks.jsx index db4604f95..4580c90f6 100644 --- a/docs/src/snippets/configs/deepseek-ai/deepseek-v4-benchmarks.jsx +++ b/docs/src/snippets/configs/deepseek-ai/deepseek-v4-benchmarks.jsx @@ -653,8 +653,18 @@ export const benchmarks = [ accuracy: { mmmu_pro_pct: 75.14 }, notes: "MMMU-Pro (standard, 10-option) measured with sgl-eval on 4×B200 (TP=4) at temperature 1.0, top-p 0.95, --reasoning-effort max, with the bundled DSpark head enabled (--speculative-algorithm DSPARK).", }, - { match: { hw: "b200", variant: "flash-vision", quant: "fp4", strategy: "balanced", nodes: "single" } }, - { match: { hw: "b200", variant: "flash-vision", quant: "fp4", strategy: "high-throughput", nodes: "single" } }, + { + match: { hw: "b200", variant: "flash-vision", quant: "fp4", strategy: "balanced", nodes: "single" }, + sglang_version: "PR #37253 @ 31854c3", + accuracy: { mmmu_pro_pct: 74.10 }, + notes: "MMMU-Pro (standard, 10-option) measured with sgl-eval on 4×B200 (TP=4, DP=4, DeepEP) at temperature 1.0, top-p 0.95, --reasoning-effort max; target-only (DP Attention is incompatible with DSpark).", + }, + { + match: { hw: "b200", variant: "flash-vision", quant: "fp4", strategy: "high-throughput", nodes: "single" }, + sglang_version: "PR #37253 @ 31854c3", + accuracy: { mmmu_pro_pct: 73.76 }, + notes: "MMMU-Pro (standard, 10-option) measured with sgl-eval on 4×B200 (TP=4, DP=4, MegaMoE) at temperature 1.0, top-p 0.95, --reasoning-effort max; target-only (DP Attention is incompatible with DSpark).", + }, // B300 / GB200 / GB300 / H200 / H100 — Flash Vision (Exp), all pending { match: { hw: "b300", variant: "flash-vision", quant: "fp4", strategy: "low-latency", nodes: "single" } }, { match: { hw: "b300", variant: "flash-vision", quant: "fp4", strategy: "balanced", nodes: "single" } }, diff --git a/docs/src/snippets/configs/deepseek-ai/deepseek-v4.jsx b/docs/src/snippets/configs/deepseek-ai/deepseek-v4.jsx index 8111c2e63..88821e0bb 100644 --- a/docs/src/snippets/configs/deepseek-ai/deepseek-v4.jsx +++ b/docs/src/snippets/configs/deepseek-ai/deepseek-v4.jsx @@ -3047,8 +3047,7 @@ sgl-eval run mmmu_pro \\ }, { match: { hw: "b200", variant: "flash-vision", quant: "fp4", strategy: "balanced", nodes: "single" }, - verified: false, - verificationStatus: "in-progress", + verified: true, warn: "DeepSeek-V4-Flash-Vision-Exp support has not shipped in an SGLang release yet (sglang PR 37253): Docker mode already points at the preview image; for Python mode install SGLang from that PR. See [Flash Vision notes](#vision-note).", env: ["SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=1024"], flags: [ @@ -3065,8 +3064,7 @@ sgl-eval run mmmu_pro \\ }, { match: { hw: "b200", variant: "flash-vision", quant: "fp4", strategy: "high-throughput", nodes: "single" }, - verified: false, - verificationStatus: "in-progress", + verified: true, warn: "DeepSeek-V4-Flash-Vision-Exp support has not shipped in an SGLang release yet (sglang PR 37253): Docker mode already points at the preview image; for Python mode install SGLang from that PR. See [Flash Vision notes](#vision-note).", env: [ "SGLANG_OPT_DEEPGEMM_MEGA_MOE_NUM_MAX_TOKENS_PER_RANK=8320",