Search results

[Bug] MiniMaxAI/MiniMax-M2.5 on 8xA100 (TP=8, EP=8) fails on Ampere (both Marlin and Triton)

Status: Open.

#18875 In sgl-project/sglang;

· ehartford opened on Feb 16, 2026

[Bug] GLM-4.7-Flash produces garbage output on NVIDIA Blackwell (sm_120) GPUs

Status: Open.

#18874 In sgl-project/sglang;

· realtmxi opened on Feb 16, 2026

Document & auto-enable FP8 block-wise CUTLASS GEMM for SM120

documentationImprovements or additions to documentationImprovements or additions to documentation

good second issueGood for second time and more contributorsGood for second time and more contributors

Status: Open.

#18872 In sgl-project/sglang;

· b8zhong opened on Feb 15, 2026

[Benchmark] MiniMax-M2.5 FP8 on 8× RTX PRO 6000 (SM120): bench_serving results + tuned fused_moe_triton config

Status: Open.

#18870 In sgl-project/sglang;

· stewtong opened on Feb 15, 2026

[Bug] Failing to import <code>deep_gemm</code>

Status: Open.

#18852 In sgl-project/sglang;

· mmangkad opened on Feb 15, 2026

[Feature][NPU] Support Kimi-K2.5 on Ascend NPU

Status: Open.

#18841 In sgl-project/sglang;

· husf1130 opened on Feb 14, 2026

[Bug] Qwen3-next mtp topk&gt;1 accuracy issue

Status: Open.

#18831 In sgl-project/sglang;

· hanming-lu opened on Feb 14, 2026

[Bug] GLM5-FP8 enabling context parallel and MTP simultaneously causes error

Status: Open.

#18830 In sgl-project/sglang;

· kkkk1nak0 opened on Feb 14, 2026

[Bug] KeyError: 'ministral3'

Status: Open.

#18819 In sgl-project/sglang;

· anttttti opened on Feb 13, 2026

[Feature] Solution to Broken Documentation Link Checks

Status: Open.

#18816 In sgl-project/sglang;

· shljessie opened on Feb 13, 2026

[Bug] RotaryEmbedding fallback requires CUDA_HOME on HIP

Status: Open.

#18812 In sgl-project/sglang;

· alphabetc1 opened on Feb 13, 2026

[Feature] Support video input for Kimi K2.5

Status: Open.

#18798 In sgl-project/sglang;

· zyuhan1999 opened on Feb 13, 2026