Skip to contentT/TTokens or Towers

Compare

LeaderboardEvery tracked model, ranked and priced side by sideBenchmarksWhat each test measures and who leads itPricesAPI, media, and hardware rates with a cost calculator

Plan

Stack plannerSix questions to a priced local and hosted stackShortlistOur picks for subscriptions, APIs, and hardwareEnterpriseSeat plans against self-hosting for a whole team

Build

HardwareGPUs and workstations, and which models they fitHostsWhere to rent open-weight inference, and for how muchHarnessesCoding agents and the models they drive

Lab

FindingsResults from our own evaluationsScanAudit a repo’s AI usage in the browserMethodWhere the numbers come from and how we check them
Data checked September 25, 2026
Data checked September 25, 2026
Plan a stack
T/TTokens or Towers

An independent guide to AI models, what they cost, and the hardware to run them. No account, no API key, no sponsors.

Compare

  • Leaderboard
  • Benchmarks
  • Prices

Plan

  • Stack planner
  • Shortlist
  • Enterprise

Build

  • Hardware
  • Hosts
  • Harnesses

Lab

  • Findings
  • Scan
  • Method
Data checked September 25, 2026 · Every number links to its sourceRSS · Source on GitHub
Leaderboard/NVIDIA
NVIDIA

NVIDIA · Nemotron 3.5

Nemotron 3.5 Lightning 30B A3B

Open weightsGenerally availableNewOpenMDW-1.1

Small open hybrid Mamba-MoE (30B total / 3B active) tuned for fast, high-volume agent tasks and for customization on a single GPU.

AA Intelligence
12.9#85 of 142
Input / output per 1M
$0.07/ $0.22
Output speed
264tok/s
Context window
1M tokens

Composite indexes

  • Artificial Analysis Intelligence Index#85 of 142
    13independentIntelligence Index v4.3; reasoningArtificial Analysis: Nemotron 3.5 Lightning ↗

    Leader: Claude Opus 5.5 at 58

Coding

  • SciCode#85 of 94
    32.6%vendorNVIDIA Nemotron 3.5 Lightning model card (NVIDIA-measured comparison table) ↗
    32.1%independentreasoningArtificial Analysis: Nemotron 3.5 Lightning ↗

    Leader: Claude Opus 5.5 at 66.9%

  • Terminal-Bench 4.0#56 of 88
    0.5%independentreasoningArtificial Analysis: Nemotron 3.5 Lightning ↗

    Leader: Claude Mythos 5.1 at 60.9%

  • Terminal-Bench 2.x#80 of 103
    24.6%vendorTerminal-Bench 2.1NVIDIA Nemotron 3.5 Lightning model card (NVIDIA-measured comparison table) ↗
    24.3%independentTerminal-Bench 2.1; reasoningArtificial Analysis: Nemotron 3.5 Lightning ↗

    Leader: Claude Fable 5.1 at 91.4%

  • SWE-bench Verified#13 of 16
    51.6%vendorNVIDIA Nemotron 3.5 Lightning model card (NVIDIA-measured comparison table) ↗

    Leader: Claude Sonnet 5 at 85.2%

Agents and tool use

  • GDPval-AA#70 of 103
    832vendorGDPval-AA v2NVIDIA Nemotron 3.5 Lightning model card (NVIDIA-measured comparison table) ↗
    624independentreasoningArtificial Analysis: Nemotron 3.5 Lightning ↗

    Leader: Claude Opus 5.5 at 1846

  • BrowseComp#13 of 14
    37.0%vendorNVIDIA Nemotron 3.5 Lightning model card (NVIDIA-measured comparison table) ↗

    Leader: GPT-6 Astra at 91.5%

Reasoning and knowledge

  • GPQA Diamond#91 of 142
    75.4%vendorno toolsNVIDIA Nemotron 3.5 Lightning model card (NVIDIA-measured comparison table) ↗
    74.3%independentreasoningArtificial Analysis: Nemotron 3.5 Lightning ↗

    Leader: GPT-6 Astra at 96.1%

  • Humanity’s Last Exam#102 of 147
    11.7%vendortext-only, no toolsNVIDIA Nemotron 3.5 Lightning model card (NVIDIA-measured comparison table) ↗
    10.6%independentno tools; reasoningArtificial Analysis: Nemotron 3.5 Lightning ↗

    Leader: Claude Opus 5.5 at 61.4%

  • CritPt#110 of 139
    0%independentreasoningArtificial Analysis: Nemotron 3.5 Lightning ↗

    Leader: GPT-5.6 Sol at 32.3%

  • MMLU-Pro#6 of 14
    81.9%vendorNVIDIA Nemotron 3.5 Lightning model card (NVIDIA-measured comparison table) ↗

    Leader: Nemotron 3 Ultra 550B A55B at 86.8%

Vision and long context

  • AA Long Context Reasoning#84 of 139
    60.3%independentreasoningArtificial Analysis: Nemotron 3.5 Lightning ↗
    52%vendorNVIDIA Nemotron 3.5 Lightning model card (NVIDIA-measured comparison table) ↗

    Leader: Kimi K3 at 88.7%

Other published results

ResultScoreReported bySource
tau2-Bench Banking (AA) · reasoning8.9%independentArtificial Analysis: Nemotron 3.5 Lightning ↗
AA-Omniscience Index · reasoning-17.7 indexindependentArtificial Analysis: Nemotron 3.5 Lightning ↗
PinchBench85.37%vendorNVIDIA Nemotron 3.5 Lightning model card (NVIDIA-measured comparison table) ↗
SWE-bench Multilingual39.33%vendorNVIDIA Nemotron 3.5 Lightning model card (NVIDIA-measured comparison table) ↗
IFBench · loose71.88%vendorNVIDIA Nemotron 3.5 Lightning model card (NVIDIA-measured comparison table) ↗

Specs

Released
August 11, 2026
Parameters
30B total, 3B active
Size class
10 to 40B
Input
text
Output
text
Context
1M tokens
Max output
Not published
Cached input
$0.05 / 1M
Blended 70/30
$0.115 / 1M
First token
0.60s
  • Price: Artificial Analysis model page ↗ No first-party paid NVIDIA API (build.nvidia.com is a free trial endpoint). Value is the Artificial Analysis median across third-party API providers.
  • Speed: Artificial Analysis leaderboard (Nemotron 3.5 Lightning, median across providers or first-party as shown) ↗

Will it fit your budget?

The planner prices this model against local hardware and other APIs for your workload.

Plan a stack →

Compare with

  • NVIDIANemotron 3 Super 120B A12B12.8
  • CohereCommand A+13.1
  • AmazonAmazon Nova 2 Lite13.4
  • ServiceNowApriel v1.6 15B Thinker13.4