Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add NVIDIA-NeMo/Megatron-Bridge --skill nemo-mbridge-perf-moe-dispatcher-selectiongit clone --depth 1 https://github.com/NVIDIA-NeMo/Megatron-BridgeWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/nvidia-nemo/megatron-bridge/nemo-mbridge-perf-moe-dispatcher-selection)<a href="https://agentmods.dev/skills/nvidia-nemo/megatron-bridge/nemo-mbridge-perf-moe-dispatcher-selection"><img src="https://agentmods.dev/badge/skills/nvidia-nemo/megatron-bridge/nemo-mbridge-perf-moe-dispatcher-selection/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/nvidia-nemo/megatron-bridge/nemo-mbridge-perf-moe-dispatcher-selection"><img src="https://agentmods.dev/badge/skills/nvidia-nemo/megatron-bridge/nemo-mbridge-perf-moe-dispatcher-selection.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00076 | $0.02346 |
| Opus 5 | $0.00038 | $0.01173 |
| Sonnet 5 | $0.00015 | $0.00469 |
| Haiku 4.5 | $0.00008 | $0.00235 |
Grade A, and why
nemo-mbridge-perf-moe-dispatcher-selection scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 232 lines — stays where its author put it; the contents beside it link to each section on GitHub.
MoE Dispatcher Selection Guide
Stable docs: @docs/training/moe-optimization.md Card: @skills/nemo-mbridge-perf-moe-dispatcher-selection/card.yaml
Quick Decision
By hardware
| Hardware | Bring-up path | Tuned candidates |
|---|---|---|
| H100 | alltoall |
A/B DeepEP and HybridEP when installed; the current 16×H100 Qwen3 30B winner is HybridEP |
| B200 | alltoall |
A/B DeepEP and HybridEP when supported by the target runtime |
| GB200 / GB300 NVL72 | alltoall |
HybridEP is the strongest topology-informed candidate; compare DeepEP when available |
| Unknown | alltoall |
Add one flex backend only after the correctness baseline is stable |
Hardware narrows the candidate set; it does not select the winner. Hold the model, routing, batch shape, parallelism, overlap, graph scope, container, and timing window fixed during the comparison.
By EP degree
| EP size | Guidance |
|---|---|
| Small EP | Dispatcher choice may be second-order; start with alltoall |
| Medium EP | Profile first, then A/B the installed flex backends |
| Large EP | Prioritize topology-aware candidates, but still require a matched A/B |
On one NVL8 domain in BF16, treat alltoall and HybridEP as matched candidates:
their throughput can be close once the full stack is held fixed. HybridEP is a
high-priority tuning path, not a reason to skip the correctness baseline.
Model-Family Patterns
| Workload | Common best path | Notes |
|---|---|---|
| DSV3 at large scale | Measured snapshots use HybridEP on GB200/GB300 and DeepEP on H100 | Revalidate against the target container and topology |
| Qwen3 235B | Current H100 recipe uses alltoall plus overlap; measured GB200 snapshots use HybridEP |
Do not replace the current recipe from a hardware rule alone |
| Qwen3 30B | Current canonical 16×H100 recipe uses HybridEP | Direct counterexample to H100 → DeepEP mapping |
| Qwen3-Next | Workload-dependent | Precision, memory, PP layout, and kernels can change the ordering |
| MoE VLMs | Start simple, then test HybridEP on GB200-class systems | Vision workloads are sensitive to both memory and host overhead |
What ships with it
5 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 12d ago First seen · 232 lines · 76 tokens per session scan A 2a6fb1c401e4
nemo-mbridge-perf-moe-dispatcher-selection is a skill published in the GitHub repository NVIDIA-NeMo/Megatron-Bridge (906 stars, last pushed today), licensed Apache-2.0. It adds 76 tokens to every session and 2,346 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
cli-eval
Create and run evaluation suites, watch live benchmark progress, view scorecards, compare model performance, and integrate eval runs with CI workflows from the CLI.
model-merging
Merge multiple fine-tuned models using mergekit to combine capabilities without retraining. Use when creating specialized models by blending domain-specific expertise (math + coding + chat), improving performance beyond single models, or experimenting rapidly with model variants. Covers SLERP, TIES-Merging, DARE, Task…
merger-model
Build accretion/dilution analysis for M&A transactions. Models pro forma EPS impact, synergy sensitivities, and purchase price allocation. Use when evaluating a potential acquisition, preparing merger consequences analysis for a pitch, or advising on deal terms. Triggers on "merger model", "accretion dilution", "M&A…
darwinian-evolver
Evolve prompts/regex/SQL/code with Imbue's evolution loop.
validate
Validate Semantica pipelines, extraction quality, graph schemas, and ontology consistency. Returns structured error/warning checklists. Uses PipelineValidator, PipelineBuilder.validatepipeline(), GraphValidator, and OntologyValidator. Sub-commands: pipeline, step, dependencies, extraction, graph, ontology, performance.
launching-evals
Run, monitor, analyze, and debug LLM evaluations via nemo-evaluator-launcher. Covers running evaluations, checking status and live progress, debugging failed runs, exporting artifacts and logs, and analyzing results. ALWAYS triggers on mentions of running evaluations, checking progress, debugging failed evals…