README: GLM-5.3-Flash from NVIDIA's FP4-Local and from XSFP4 in Pick a model

README.md

@@ -170,6 +170,8 @@ Conversions are explicitly marked: their upstream weights need the linked conver
 | GLM-5.3-Flash, MLX 4-bit | `tensorfold pull TensorFold/GLM-5.3-Flash-MLX-4bit-MTP` | Two M5 Ultras; single-Mac native serving is not qualified. | [Paired Metal runbook](RUNBOOK.md#flash-next-and-paired-metal-serving). Do not use the Python recipe's single-Mac command. |
 | GLM-5.3-Flash, EXL3/TR3 | `tensorfold pull Mia-AiLab/GLM-5.3-Flash-EXL3-TR3-4bpw` | Two CUDA ranks; the release checks the two-GB10 setup. | [Release scope](CHANGELOG.md#110), [checkpoint](docs/recipes/glm-5.3-flash.md#exl3). Experimental checkpoint; use [native rank options](docs/recipes/glm-5.3-flash-cuda.md#run), not the Python installation. |
 | GLM-5.3-Flash, NVFP4 | `tensorfold pull nvidia/GLM-5.3-Flash-NVFP4@423acf37583782c51c142d145aef733d72943d93` | Two GB10s at TP=2: @Bizuayeu loaded the whole checkpoint on 1.1.0 in 302.1 s and served it. | [Native CUDA](docs/recipes/glm-5.3-flash-cuda.md); [contributed run](https://github.com/ashhart/TensorFold/issues/1040): one token hash per prompt across three runs and a passphrase found in 199,307 tokens; scored prompts and more than two ranks were not covered. |
+| GLM-5.3-Flash, FP4-Local | `tensorfold pull nvidia/GLM-5.3-Flash-FP4-Local@76348ddb2be2e35e39ecfc00e75dc07703f4f3d5` | Two GB10s at TP=2, or three at TP=3 over a ring of direct links: @Bizuayeu loaded it as stored and served it at both sizes. 165.3 GiB of weights, 13% less than NVFP4's 190.4 GiB. | [Native CUDA](docs/recipes/glm-5.3-flash-cuda.md), on main after 1.1.4 (#1050). NVIDIA's mixed-precision checkpoint: NVFP4 routed experts with compressed block scales, MXFP8 and 128×128-block FP8 for most of the rest. As close to BF16 as NVFP4 at its smaller size: at TP=2 its NLL is 1.11% above BF16's against NVFP4's 1.37% ([contributed receipt](docs/receipts/glm-fp4local-80517fc3/README.md) on #1050's head, not rerun on main), a difference the per-document intervals in [XSFP4's card](https://huggingface.co/Bizuayeu/GLM-5.3-Flash-XSFP4#measured) do not resolve. |
+| GLM-5.3-Flash, XSFP4 | `tensorfold pull Bizuayeu/GLM-5.3-Flash-XSFP4@88736f916f8939607fde8a4b881dcae749f7551d` | Two GB10s at TP=2, or three at TP=3 over a ring of direct links: on main `8086c205c` @Bizuayeu loaded it in 370.7 s at TP=2 (rank 0's model 96.97 GiB at a 524,288-token window) and in 313.8 s at TP=3 (63.60 GiB at 1,048,576), and both served. 165.7 GiB of weights, 0.4 GiB more than FP4-Local's. | [Native CUDA](docs/recipes/glm-5.3-flash-cuda.md), on main after 1.1.4 through FP4-Local's path. FP4-Local with its routed experts' NVFP4 blocks fitted again from the BF16 weights, a smoothing folded into them, so it runs on the same kernels. Closer to BF16 than FP4-Local and NVFP4 at TP=2 by all three measures in [its card](https://huggingface.co/Bizuayeu/GLM-5.3-Flash-XSFP4#measured), on #1050's code: top-1 agreement 93.26%, NLL 0.85% above and a KL bound of 0.0431, against FP4-Local's 92.82%, 1.11% and 0.0510. |
 | GLM-5.3, EXLR8 `home` | `tensorfold pull kindlingai/glm-5.3-exlr8-k2-k3-k4@f459f5d0272a529403dd9b3fd32f771f4250a0d4 --exlr8 home --node 0/2 --dir ~/models/glm-5.3-exlr8-home` | Two M5 Ultra Macs, 256 GB each; about 160 GB resident per Mac at 8k context / four streams. | [Two-Mac recipe](docs/glm-5.3-exlr8.md). Use `--node 1/2` on the other Mac; Thunderbolt 5 and MCDMA required. No CUDA backend. |
 | LFM2.5-24B-A2B, MLX 4 / 6 / 8-bit | `tensorfold pull LiquidAI/LFM2.5-24B-A2B-MLX-4bit@667408d494bde191c2cb9dd4f4549827816b9275`<br>`tensorfold pull LiquidAI/LFM2.5-24B-A2B-MLX-6bit@1deec53eb49571e21cdcbf82fd5839248f07ed88`<br>`tensorfold pull LiquidAI/LFM2.5-24B-A2B-MLX-8bit@67ab417dd02d0c9545802a726df668620eae049d` | One Mac; M4 Pro checked. | [Native operator recipe](docs/recipes/README.md), [release history](CHANGELOG.md). No separate native model recipe yet. |
 | Qwen3.6-35B-A3B, MLX with MTP | `tensorfold pull TensorFold/Qwen3.6-35B-A3B-MLX-4bit-MTP@81169a9bc511a27c1b4eedb77a2cd98ced431847` | One GB10; 20.9 GB checkpoint. | [Qwen3.6](docs/recipes/qwen3.6-moe.md). Native CUDA; the recipe's Metal path belongs to Python. |

docs/recipes/glm-5.3-flash-cuda.md

@@ -1,10 +1,11 @@
 # GLM-5.3-Flash on CUDA
 
 The native CUDA build serves GLM-5.3-Flash at tensor parallel over hosts: from NVIDIA's ModelOpt NVFP4 checkpoint
-([1.0.12](../../CHANGELOG.md#1012)), and on two ranks from Mia-AiLab's EXL3/TR3 checkpoint
-([1.1.0](../../CHANGELOG.md#110), [the checkpoint](glm-5.3-flash.md#exl3)). Rank 0 serves HTTP and the other ranks
-follow its steps, with MTP drafts, concurrent streams and kept prompt states. The
-[Python 0.6.6 recipe](glm-5.3-flash.md) and its `--tp`, `--rank` and `--master` flags belong to the frozen engine.
+([1.0.12](../../CHANGELOG.md#1012)), on two ranks from Mia-AiLab's EXL3/TR3 checkpoint
+([1.1.0](../../CHANGELOG.md#110), [the checkpoint](glm-5.3-flash.md#exl3)), and on main after 1.1.4 (#1050) from
+NVIDIA's FP4-Local checkpoint as stored and from XSFP4, which keeps FP4-Local's tensors outside the routed experts.
+Rank 0 serves HTTP and the other ranks follow its steps, with MTP drafts, concurrent streams and kept prompt states.
+The [Python 0.6.6 recipe](glm-5.3-flash.md) and its `--tp`, `--rank` and `--master` flags belong to the frozen engine.
 
 ## Run