Fast NVIDIA Nemotron MoE for reliable agentic tasks across enterprise workloadsNemotron 3.5 Lightning 30B A3B by nvidia is offered by 21 providers on this page. Public prices are shown for 12 of them. Its official list price is $0 per 1M tokens. No paid channel is priced below the official list price. Prices vary about 7.4× across the providers that host it. It also has 9 free ($0) channels; free tiers usually carry rate limits, and subscription-covered access bills $0 per token only after the subscription fee.
Artificial Analysis rates it 5.5 on the Intelligence Index, with 15.5 for Engineering, its strongest domain. Against its lowest blended price of $0.049 per 1M, that is roughly 113 index points per dollar, which is the value ratio the leaderboards rank on. Median output speed is 297 tokens per second, with 0.63s to the first token. Running one task of the Artificial Analysis suite costs about $0.3145, which reflects how many tokens its reasoning consumes rather than the unit price alone.
The context window is 262,144 tokens at the reference host, but hosts report different limits, from 65,536 to 1,048,576, so the usable window depends on the provider you pick. It supports reasoning, tool use, and structured output. The weights are open, so it can also be self-hosted or served through a gateway of your choice. Providers report the capability flags inconsistently, so verify a specific feature against the host you plan to use. Its training knowledge cuts off at 2025-09.
Quality is independently evaluated by Artificial Analysis. Speed/latency are model-level medians.
| Provider | Tier | Input | Output | Cache read | Cache write | Context | Output limit | Status |
|---|---|---|---|---|---|---|---|---|
| NvidiaOfficial nvidia/nemotron-3.5-lightning-30b-a3b | Cloud | Free | — | — | 262,144 | 262,144 | ||
| Requesty nemotron-3.5-lightning-30b-a3b | Gateway | Free | — | — | 1,048,576 ⚠ | 65,536 | ||
| Merge Gateway nvidia/nemotron-3.5-lightning-30b-a3b | Gateway | Free | — | — | 1,000,000 ⚠ | 262,144 | ||
| OpenRouter nvidia/nemotron-3.5-lightning:free | Gateway | Free | — | — | 1,000,000 ⚠ | 65,536 | ||
| AIHubMix nemotron-3.5-lightning-free | Gateway | Free | — | — | 262,144 | 262,144 | ||
| Kilo Gateway nvidia/nemotron-3.5-lightning:free | Gateway | Free | — | — | 1,000,000 ⚠ | 65,536 | ||
| InferAll nvidia/nemotron-3.5-lightning-30b-a3b | Gateway | Free | — | — | 262,144 | 262,144 | ||
| Vercel AI Gateway nvidia/nemotron-3.5-lightning-free | Cloud | Free | — | — | 1,000,000 ⚠ | 65,000 | ||
| OpenCode nemotron-3.5-lightning-free | Gateway | Free | $0 | — | 262,144 | 262,144 | ||
| Kilo Gateway | Gateway | $0.025 | $0.12 | $0.013 | — | 262,144 | 235,929 |
Sorted by blended price (input×0.75 + output×0.25) asc. The official channel always shows regardless of rank. Whether a gateway's low price is actually usable can't be verified.
Interleaved thinking (reasoning between tool calls) is declared by 4 of 21 providers.
1 / 21 providers expose no reasoning control (reasoning_options: []).
9 more channels offer $0 (Nvidia, Requesty, Merge Gateway etc.); free tiers usually have rate limits and no SLA, excluded from ranking.
| Name | Conditions | Score | Metric | Source |
|---|---|---|---|---|
| MMLU-Pro | harness: NeMo Gym / NeMo Evaluator SDK · variant: BF16; reasoning | 81.94 | score | Source ↗ |
| AA-Omniscience | harness: NeMo Gym / NeMo Evaluator SDK · variant: BF16; reasoning | 17.5 | score | Source ↗ |
| SciCode | harness: NeMo Gym / NeMo Evaluator SDK · variant: BF16; reasoning | 32.6 | score | Source ↗ |
| PinchBench | harness: NeMo Gym / NeMo Evaluator SDK · variant: BF16; reasoning | 85.37 | score | Source ↗ |
| BrowseComp | harness: NeMo Gym / NeMo Evaluator SDK · variant: BF16; reasoning | 36.97 | score | Source ↗ |
| AA-LCR | harness: NeMo Gym / NeMo Evaluator SDK · variant: BF16; reasoning | 52 | score | Source ↗ |
| GPQA Diamond | variant: BF16; reasoning; no tools | 75.44 | accuracy | Source ↗ |
| Humanity's Last Exam | variant: BF16; reasoning; no tools · dataset: text-only | 11.72 | accuracy | Source ↗ |
| SWE-Bench Verified | harness: NeMo Evaluator · variant: BF16; reasoning | 51.56 | accuracy | Source ↗ |
| SWE-Bench Multilingual | harness: NeMo Evaluator · variant: BF16; reasoning | 39.33 | accuracy | Source ↗ |
| Terminal-Bench | harness: NeMo Evaluator · variant: BF16; reasoning · v2.1 | 24.58 | accuracy | Source ↗ |
| tau3-bench | variant: BF16; reasoning; banking | 9.28 | score | Source ↗ |
| GDPval-AA | variant: BF16; reasoning · v2 | 832 | Elo | Source ↗ |
| IFBench | variant: BF16; reasoning; loose | 71.88 | accuracy | Source ↗ |
The same benchmark scores very differently across harness / dataset, so the qualifying conditions must be shown together.
Individual evaluations run by Artificial Analysis, on the same reasoning tier as the intelligence score above. Each benchmark has its own task set and scale (percent correct, points or an index), so rows are not comparable with one another.